Swedroe-type portfolio: downside risk vs tracking error risk

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Swedroe-type portfolio: downside risk vs tracking error risk

Post by Robert T » Tue Dec 17, 2013 5:38 am

.
Larry has often said that increasing a portfolio tilt to small and value stocks, while reducing beta (equity exposure) can help achieve the same ‘expected’ return but with lower volatility (reduced tail risk) relative to a market portfolio. This helped significantly reduce the downside in 2000-02 and 2008. His own portfolio is structured in a similar way. In addition, Bill Bernstein in “The Ages of the Investor” also presents a portfolio with a significant value and small cap tilt for the young investor.

I think there is some merit in a factor tilted approach, but with a big caveat – tracking error risk.

This post is to highlight the magnitude of tracking error risk, and possible ways to address some of it for those wanting to consider a factor tilted approach.

So where’s the catch? Tracking error risk
  • The data from 1982-2012 illustrate the downside benefit. A 100% holding of S&P500 had an annualized return of 11.1 %, with cumulative losses in 2000-2002 of -37.6%, and in 2008 of -37.0%. A 50% RAFI pure small value: 50% T-bills portfolio had a similar back-tested annualized return of 11.2%, but had a cumulative gain in 2000-2002 of 29%, and a much lower downside in 2008 of -17.9%. So where's the catch? In my view it is tracking error risk – which can be huge and for long periods. For those considering this approach, I would suggest considering tracking error risk before taking the plunge. If we rewind to 1984, and compare performance of the above mentioned portfolios from 1984 to 1999.
    • 1984-1999 – growth in $1.
      S&P500 = $14.3
      50% RAFI Pure Small Value: 50% T-bills = $5.4
    By 1999 the tilted portfolio would have been 37% of the size of the S&P500 portfolio (i.e. the S&P500 portfolio would have been more than two and half times the size of the tilted portfolio). How many people do you know could stick with this tilted strategy approach for 15 years of under-performance (among professionals, clients, and individual investors)? Honestly. My sense is very few.

    Obviously this underperformance at a portfolio level is lower if an investor only has a small equity allocation (eg. 25% equities), but for those in the accumulation stage with higher equity allocation, the underperformance can be significant. It is probably as hard to stick with a strategy when everyone else is outperforming for such a long period (1984-1999), than it is when everyone loses money in a crash (2008). In this respect the greatest risk with these strategies may be tracking error risk (capitulation). While the same case could be made for underperforming a value and small cap tilted portfolio, most people in the broader finance community, right or wrong, measure performance against beta (the market), and underperformance against the market may be harder to stick with, as there would likely be more noise around it. There’s a trade-off then in this case between downside risk, and tracking error risk. Greater downside protection came with greater tracking error risk (which can be significant).
What can be done to reduce tracking-error? Consider adding a large cap momentum ETF
  • Two options are:
    • 1. Add more S&P500 back to the tilted portfolio. This would both increase the downside a bit, but also reduce tracking error. Stop adding when you find the balance between downside protection and tracking error risk you can live with – both are important (some prefer lower beta e.g. Larry Swedroe, some prefer higher beta e.g. Rick Ferri – no right answer for everyone).

      2. An alternative is to consider adding a large cap momentum fund (in place of adding back S&P500) – which tends to amplify returns when the S&P500 does well as in 1984-1999, off-setting more of the lag in small value. In addition, the low average correlation between momentum and value (Asness) can add a diversification benefit.
    On the second approach, one option is the recently opened iShares MSCI Momentum ETF. Using data from Jan 1982-Oct. 2013, the momentum load on the MSCI momentum index is a bit lower than the AQR momentum index (0.28 vs. 0.37), but it seems to avoid the negative ‘profitability’ load of the AQR index (+0.09 vs. -0.14). It has a zero value load, and is large cap (-0.18 size load). Just to note that the momentum load on the AQR (large cap) momentum index is higher than on the AQR small cap momentum index (0.37 vs. 0.28) perhaps suggesting it is easier to capture the momentum premium with large caps than small caps. In addition turnover of momentum funds are typically higher, which may be more cost effective to implement with large caps stocks than small caps stocks (liquidity, bid-ask spreads, market impact).
    • Annualized return 1982-2012
      MSCI momentum index = 13.5%
      AQR momentum index = 12.8%
      S&P500 = 11.1%

      Correlation coefficient between MSCI momentum index and the RAFI fundamental pure small value index = 0.41, and between S&P500 and RAFI fundamental pure small value = 0.62.
    Over the basktest period 1982-2012, adding the MSCI momentum index to the SV tilted portfolio did not cost much in terms to lower downside risk (-21.2% vs -17.9% in 2008), but gained in lower tracking error risk in that the portfolio with momentum added, gained 40 percent more in portfolio value from 1982-99 than the portfolio with only SV and T-bills i.e. less underperformance relative to a 100% S&P500 portfolio. However, the portfolio with momentum added, was still only 52% of the value of the S&P500 portfolio by 1999 – so still significant tacking error. Simply adding back the S&P500 was less effective than adding the momentum index (lower sharpe ratio: 0.61 vs 0.68).

    Just to note that adding the MSCI quality index to the mix did not add much. This may be due to the fact that the RAFI Pure SV already has a positive load to “profitability” whether using the Novy-Marx profitability factor or the Asness, Frazzini, Pedersen “quality-minus-junk” factor. This seems a bit counterintuitive, but perhaps its the result of weighting/sorting of stocks by cash flow and sales (re: Larry’s earlier blog on this). The MSCI momentum fund also has a small but positive profitability load. So using MSCI momentum and RAFI pure SV in a portfolio already provides a significant portfolio "profitability load" and in this sense adding the MSCI quality index does not add much to this, but subtracts from the portfolio value tilt as the MSCI quality index had a -0.15 value load.

    Image
Bottom line
  • The bottom line seems to be that heavily tilting a portfolio to small and value stocks while reducing beta has reduced downside risk relative to a market portfolio without giving up much return over 1982-2013. The cost was significant tracking error in 1984-1999 – needing some fairly incredible staying power. For those wanting less tracking error risk, adding a momentum fund would have reduced this more effectively than simply adding back the S&P500 (resulting in a higher Sharpe ratio).
My own preference if for a higher beta portfolio, with less than a full SV tilt given my tolerance to tracking error risk (know thyself – not sure that I would have lasted 1984-1999).

Just some due diligence on momentum, now that investable options are available. Hope its useful.

Robert
.

User avatar
matjen
Posts: 2021
Joined: Sun Nov 20, 2011 11:30 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by matjen » Tue Dec 17, 2013 7:41 am

Robert thank you for this wonderful analysis. Larry's portfolio or something similar is potentially the next strategy that I may employ for my own. Tax bills and tracking error and fear of being a "tourist" (as Rick F. cleverly puts it) have kept me from doing anything. The great bull run this year for a high equity and tilted portfolio hasn't hurt either! Is there a reason you didn't just do this analysis with Larry's recommended approach which has more geographic diversity? Int'l SCV and EM value? I assume that doesn't change much of your overall conclusion/thoughts?

Thanks again for the great post.
A man is rich in proportion to the number of things he can afford to let alone.

User avatar
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Rick Ferri » Tue Dec 17, 2013 7:48 am

Thanks Robert. I believe it's prudent to remind readers that no investor actually earned the returns from your analysis. The value and momentum returns are hypothetical using indexes that were created with the benefit of 20/20 hindsight. The funds in your analysis did not exist to investors during most of the time period you analyzed. A RAFI Pure Small Value ETF has only been available for a couple of years, and momentum index funds have even less live data. Thus, the added risk of this strategy is that the hypothetical performance of the past will not be the actual performance in the future.

Rick Ferri
The Education of an Index Investor: born in darkness, finds indexing enlightenment, overcomplicates everything, embraces simplicity.

larryswedroe
Posts: 15726
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by larryswedroe » Tue Dec 17, 2013 9:04 am

Few thoughts--beyond that issue of tracking error risk which is very real but purely psychological. Only "fools" confuse strategy and outcome. Think about it this way. The Larry Portfolio strategy is like buying portfolio insurance. It insures against the big fat left tail risk with the "price" being not lower expected returns but giving up the good fat right tail opportunity. Now think about the insurance analogy. You buy life insurance and you don't die and you complain that the strategy was wrong. It's no different.

On Rick's point. There are good papers by Toby Moskowitz, etal on the trading/implementation costs of MOM strategies and they much smaller than people think and have small impact on the premiums. We also have now 20 years of data on live DFA funds and they show ability to keep costs down.

On Robert's points, adding in S&P or LV or anything with lower expected returns than SV requires then that you also add more tail risk --so that's the trade there. It's all matter of degree.
However, if you have a SV fund that screens for profitability, that will cut TE risk to some degree, on order or 1-2 percent.

Finally, Kevin Grogan and I are putting the finishing touches on a short book (about 65 pages or so) that will walk people through the science and evidence and the risks of the low beta/high tilt strategy. Should be out in first quarter.
Best wishes
Larry

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Tue Dec 17, 2013 9:32 am

.
Just to add the two period annualized returns 1982-1999 and 2000-12

S&P500
1982-1999 = 18.5%
2000-2012 = 1.7%

50% RAFI Pure SV: 50% T-bills
1982-1999=12.7%
2000-2012=9.1%

28% MSCI Momentum: 28% RAFI Pure SV: 44% T-bills
1982-1999=14.4%
2000-2012=6.8%

With the caveats of simulated data from Rick.

Matjen - have not looked at intl side in this yet.

Robert
.

garlandwhizzer
Posts: 2155
Joined: Fri Aug 06, 2010 3:42 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by garlandwhizzer » Tue Dec 17, 2013 11:31 am

Robert T's analysis is brilliant as is Larry's. Both make compelling cases for optimizing asset allocation based on past performance. However... the skeptic in me wonders.

Because a magic combination of selected indexes, SCV, momentum, etc., has optimized past investment returns in the past, how certain is it that real funds having real expenses, generating real taxes, and trying but perhaps not fully succeeding in tracking these magic index formulas will outperform going forward? It is very easy to mine index data from the past to determine what optimal asset allocation would have been, but will it continue to outperform in after tax dollars with the real funds in the future? It seems to me that that is a big step and it involves both faith and science. No argument from me that it is fine to follow such an investment path if one wishes to, but I personally am not certain that it will prove superior to 3 fund portfolio going forward in spite of what has happened in the past.

There are so many factor-based "smart beta" funds now, sucking up so much money, all trying to mine the same traditional outperforming factors. I wonder if the available factor outperformance will be overgrazed into plain beta, in which case simpler cheaper broadly based index funds might outperform simply on the basis of lower costs. Also it should be noted that in trying to combine multiple factors, there is an inherent conflict between some of them, momentum and value for instance, or profitability and value, which means that loading up on one factor (value) may necessarily reduce exposure to the others (momentum, profitability). I personally choose to do factor exposure with small and value, but only a modest tilt (25% US equity exposure) added to a large exposure (75%) of TSM. In the end, I am not certain which will outperform over a given time period, but I am certain of the Cost Matters Hypothesis.

Garland Whizzer

phositadc
Posts: 385
Joined: Mon Jul 26, 2010 6:17 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by phositadc » Tue Dec 17, 2013 1:19 pm

I strongly considered a portfolio of this type for awhile but ultimately rejected it. Although I love the idea of minimizing the fat downside tail, I know myself well enough to know that I could never maintain the asset allocation in a period like 1984-1999 which large caps are outperforming small caps. I would fold. So instead of doing something like 30% SCV / 70% treasuries, I did a more typical 50% total market / 50% bonds (much of which are treasuries). My downside risk is higher, but I'm pretty sure I'll be able to maintain this regardless of whether SCV or large cap are doing better at any given time.

So I guess for me I'd rather be mimicking the total market--whether good or bad--than risk significant underperformance for periods of time.

michaelsieg
Posts: 553
Joined: Mon Jan 07, 2013 11:02 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by michaelsieg » Tue Dec 17, 2013 2:02 pm

Thanks Robert,
I had a question about the momentum factor, how do the AQR small cap momentum and the AQR momentum index compare to the MSCI momentum index - could the AQR small cap momentum index be used to reduce tracking error the market performance (I have access to it in mo 403 plan), or would the fact that it tries to capture momentum in small caps lead to no reduction in tracking error? I just don't understand the concept of the momentum factor well enough to know, if large cap or small cap momentum moves independently or in sync with each other. Looking at the AQR website, at first glance, the performance curves seem to move fairly in sync.

Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Browser » Tue Dec 17, 2013 2:08 pm

What is the relevance of Bernstein's tradeoff between deep vs. shallow risk to the discussion? Minimize left tail allocations are more effective at controlling shallow risk (volatility), but at the obvious potential cost of a very low equity allocation which is likely to be worse at minimizing deep risk. This reminds me of another left-tail allocation strategy - the Permanent Portfolio - which is also good at controlling shallow risk at the cost of greater exposure to deep risk. Long term investors should be more concerned with deep risk than shallow risk. For them, the relative importance of tracking error vs. drawdowns favors paying more attention the the former rather than the latter.
We don't know where we are, or where we're going -- but we're making good time.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Tue Dec 17, 2013 4:19 pm

.
Garland Whizzer,
  • I can just say that I haven’t been disappointed so far.
    • 2003-2012 annualized return
      Personal portfolio (actual returns) = 10.1%* (2008 = -28.7%)
      MSCI All Cap World Index = 8.1% (2008 = -42.1%)
      US Total Bond Market = 5.2% (2008 = +5.2%)
    I also have concerns that a lot of the new products on the market with value and small cap tilts (fundamental indexes, equal weighted, low volatility etc) may simply be induced by recent strong performance of small and value tilted portfolios - actual and in backtests since 2000. And it may be the case that simple beta may be the factor that performs well again for a while. But I believe over the long-term the value and small cap premium will be positive due to their higher risk. Obviously no guarantees.
Michaelsieg,
  • The momentum load on the AQR small cap fund is 0.28 (Jan 1982 – Oct 2013), value load = 0.09, size load = 0.91, and market load = 1.06. Will post a bit more on this later.
Browser,
  • What was interesting is that both Larry and Bill Bernstein suggest an equity composition of small cap value stocks – even if Larry’s approach focuses more on ‘shallow risk’ and Bill Bernstien’s focuses more on ‘deep risk’ (for the young investor). The relevance is that investors face tracking error risk in both. Here is the earlier table replicated with 100% stocks (for the iron stomach young investors). Similar conclusions as before - less tracking error (relative to S&P500) in 1984-1999 period when momentum index was added to RAFI pure VS series. Just to note that the time horizon in the Deep Risk book for the young investor using this type of strategy is 30 years, and the benefit emphasized in the book is that the volatility of SV is higher giving more chances to buy low than S&P500 only. Adding large cap momentum index reduces this benefit, but also reduced tracking error risk. Again, this uses similated data with all the caveats pointed out by Rick.

    Image
Robert
.

User avatar
Rick Ferri
Posts: 8685
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Rick Ferri » Tue Dec 17, 2013 4:33 pm

garlandwhizzer wrote:There are so many factor-based "smart beta" funds now, sucking up so much money, all trying to mine the same traditional outperforming factors. I wonder if the available factor outperformance will be overgrazed into plain beta, in which case simpler cheaper broadly based index funds might outperform simply on the basis of lower costs.
My point exactly. This is why I recently wrote an article on the topic:

Smart Beta And Tourist Investors

"Wall Street is always coming up with cunning new marketing techniques to attract tourist investors. These are less-sophisticated individual investors and advisers who are easily wowed by glitzy industry trends, only to abandon them when the strategy falls short of expectations. This hurts their long-term returns. The latest spin to attract tourist money is smart beta.”

I'm not against multifactor strategies, in fact I personally invest this way. I am against making factor investing sound like a sure thing and Wall Street inferring that every mom and pop investor can beat the market using "smart beta" strategies. That is impossible and we all know it. Unfortunately, there are the messages being pushed with the recent marketing craze.

Rick Ferri
The Education of an Index Investor: born in darkness, finds indexing enlightenment, overcomplicates everything, embraces simplicity.

User avatar
Kevin M
Posts: 10514
Joined: Mon Jun 29, 2009 3:24 pm
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Kevin M » Tue Dec 17, 2013 6:30 pm

Thanks Robert. Very educational and interesting.

What I'm wondering is for those of us who have been tilting to small and value based on reading Larry's, Bill's, and others' books a few years ago (in which I don't believe momentum was mentioned), how seriously should we be considering adding a tilt to momentum, now that there are easily investable ways to do so? I'm talking about modest tilts, not Larry-type tilt to SV, and for those of us who basically understand and agree with the evidence on factor-based investing, have stuck with it for 5-10 years, but still harbor some skepticism as articulated by the total market crowd. I know there's no clear answer--just wondering out loud.

Kevin
Wiki ||.......|| Suggested format for Asking Portfolio Questions (edit original post)

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Wed Dec 18, 2013 6:05 am

michaelsieg wrote:I had a question about the momentum factor, how do the AQR small cap momentum and the AQR momentum index compare to the MSCI momentum index - could the AQR small cap momentum index be used to reduce tracking error the market performance (I have access to it in mo 403 plan), or would the fact that it tries to capture momentum in small caps lead to no reduction in tracking error? I just don't understand the concept of the momentum factor well enough to know, if large cap or small cap momentum moves independently or in sync with each other. Looking at the AQR website, at first glance, the performance curves seem to move fairly in sync.
Michaelsig,

Here's the above analysis repeated with the AQR small cap momentum index in place of the MSCI momentum index. On its own, the AQR small momentum index almost matched the annualized returns of the S&P500 from 1984-1999. So little tracking error over this period. And 2000-2012 performance was higher, due to the smaller cap tilt of the AQR fund. The combination with RAFI pure SV, doesn't work as well as the MSCI momentum index, but there still seems to be some benefit - lower 2000-2002 downside, not as great a tracking error if using a small value fund only in 1984-1999, and higher returns in 2000-2012. Looks okay to me. The diversification benefit seems to come from a paring of momentum with value, so perhaps in combination with a value fund (of any market cap) there could be some value added. Obviously no guarantees.

Image

Robert
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Wed Dec 18, 2013 6:22 am

Kevin M wrote:What I'm wondering is for those of us who have been tilting to small and value based on reading Larry's, Bill's, and others' books a few years ago (in which I don't believe momentum was mentioned), how seriously should we be considering adding a tilt to momentum, now that there are easily investable ways to do so? I'm talking about modest tilts, not Larry-type tilt to SV, and for those of us who basically understand and agree with the evidence on factor-based investing, have stuck with it for 5-10 years, but still harbor some skepticism as articulated by the total market crowd. I know there's no clear answer--just wondering out loud.
Kevin,

Few thoughts.

1. There is clear evidence that momentum exists - e.g.
2. The common explanation for the ‘momentum premium’ seems to be ‘behavioral-based’, although there have been some risk-based arguments put forward e.g.
  • Momentum Profits, Factor Pricing, and Macroeconomic Risk
    Momentum and Credit Ratings

    Just to note momentum has significant negative skewness – more so than any other factor and by a large margin e.g. FF momentum factor return in 2009 = -83%. And perhaps some of this higher negative skewness (higher downside risk) is priced in factor returns. Momentum seems to do poorly at early stages of significant reversals/recovery of market crashes e.g. July-August 1932 (FF momentum factor return = -73.2%), March-May 2009 (FF momentum factor return = -49.3%). These are periods when value and equity premiums were positive.

3. Momentum seems to pair well with value – e.g.

4. And there are now investible products for investors
So should investors add momentum, particularly to a value strategy? I can only speak for myself, again thinking out aloud. A common answer is that rather than adding an explicit fund focusing on momentum, chose value funds which avoid negative momentum. I think this is a reasonable approach - but what are advantages of explicitly adding momentum exposure beyond this?

The possible advantages of explicitly adding momentum:
  • 1. Diversification across more factors – which serve to reduce tracking error risk.
    • Interestingly the portfolio which gives the most even diversification across factors is a 75% MSCI Momentum: 25% RAFI pure SV. The factor loads (Jan 1982-Dec 2012) are as follows:

      1.08 = Market
      0.15 = Size
      0.20 = Value
      0.15 = Momentum
      0.14 = Profitability*

      * Using Asness “Quality minus Junk” factor. Similar result if Novy-Marx factor is used.
    2. Historically the ‘momentum premium’ has been as larger than ‘equity, size and value premiums’, if this continues then don’t need much positive exposure to get a benefit.
    • Annualized US factor premiums 1927-2012. Equity=5.9%, Size=2.6%, Value=4.1%, Momentum=7.6%
      Average US factor premiums 1927-2012. Equity=8.0%, Size=3.5%, Value=5.0%, Momentum=9.6%
    3. Low cost options are now available
Possible disadvantages
  • 1. These momentum products are new, with no proven live track record on implementation.

    2. Takes ups space in a portfolio so adding momentum fund would likely lower the value load on an overall portfolio
    • e.g. Adding the momentum fund to RAFI SV reduced the value load of the overall portfolio from 0.72 to 0.37 (-0.35), but changed the momentum load by +0.26 (from large and negative, to marginally positive).
    3. No certainties that the momentum premium will persist (if "behavioral"), then investor will be left with higher cost ‘market fund’ (higher expense ratio, trading costs etc).
Personally, don’t plan to add now. But will do a bit more due diligence, and will consider again next calendar year.

Robert
.

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by richard » Wed Dec 18, 2013 6:44 am

1) Compare Zvi Bodie's strategy - TIPS plus long-term equity options. This allows an even larger percentage of a portfolio to be in bonds, while preserving upside if equities do well.

2) Larry reduces tail risk by holding more bonds. These days we see lots of fear that bonds are unusually risky due to low rates.

3) "The Larry Portfolio strategy is like buying portfolio insurance." This may not be the best way to describe the strategy, as "portfolio insurance" is often viewed as the cause of the October 1987 crash and a failed hedging strategy.

User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by in_reality » Wed Dec 18, 2013 8:03 am

Robert T wrote:.
Bottom line
  • The bottom line seems to be that heavily tilting a portfolio to small and value stocks while reducing beta has reduced downside risk relative to a market portfolio without giving up much return over 1982-2013. The cost was significant tracking error in 1984-1999 – needing some fairly incredible staying power. For those wanting less tracking error risk, adding a momentum fund would have reduced this more effectively than simply adding back the S&P500 (resulting in a higher Sharpe ratio).
Interesting. Thanks for pointing this out.

Other research shows combining Value Strategy, Equal-Weighted Strategy, Low-Volatitily Strategy and Momentum Strategy produces a lower tracking error risk (i.e. "Active Risk") than any one strategy alone with a higher total return than the strategy with the lowest tracking error risk.

My personal portfolio-fu ability to impliment 4 strategies is surely lacking and I wish Vanguard would implement a "Composite Beta" fund incorporating all those strategies so I don't cut myself too badly on all the slicing and dicing .... :twisted: Um maybe simply going total market by market cap is what's best for someone like me... :confused

Image
http://www.indexuniverse.eu/europe/publ ... 1&start=10

larryswedroe
Posts: 15726
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by larryswedroe » Wed Dec 18, 2013 8:16 am

The key to all these strategies is that they don't correlate highly with the largest exposure people have which is BETA. If your 60/40 much more than 60% of your risk is in beta.
That provides a diversification benefit as some strategies tend to do well when others do poorly. That cuts volatility and tail risks
The simplest way to do it IMO is low beta/high tilt (and do it with funds that screen for negative MOM and use profitability type screens). But can add funds that gain direct exposure

Larry

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by richard » Wed Dec 18, 2013 10:14 am

I'm reminded of something Cliff Asness recently said, "To me, if you deviate markedly from capitalization weights, you are, by definition, an active manager making bets"
http://www.cfapubs.org/doi/pdf/10.2469/faj.v70.n1.2

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by grap0013 » Wed Dec 18, 2013 1:10 pm

It is interesting that when SCV has legged broader US markets substantially that it has historically still gained double digit returns. This is often demonstrated by Taylor Larimore with his 1984-1999 data references. By changing strategies due to this type of tracking error is often cited by Larry as, "confusing strategy with outcome."

My question is, "what is the etiology of the feelings of tracking error in these types of instances (ie SCV lags LCG substantially) ?" It is most likely greed and/or envy because you see your neighbors buying new sports cars etc.... If those emotions are your drivers for changing from a SCV tilt then you probably deserve the subpar returns you are going to get.

Just add greed/envy in addition to fear etc...to the laundry list of emotions that are hazardous to one's wealth. People who are aware that they are emotional about money are likely best served by hiring a fee based financial advisor to coldly hold their hands during particularly "emotional" times in the market.
There are no guarantees, only probabilities.

scone
Posts: 1455
Joined: Wed Jul 11, 2012 4:46 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by scone » Wed Dec 18, 2013 1:44 pm

It's not just greed and envy, it's hard-wired competitiveness. Especially, I have to say, among males who aspire to alpha status. So getting an advisor may not help, if that advisor is himself an aspiring alpha. It may just enable more risk-taking than is necessary. You see this type of hyper-competitiveness all over Wall Street, and the rewards system of the finance industry is structured to encourage it. Hence the obsession with beating the market.

It's hard for some to change their perspective from "making more money" to "losing less money," but that's partly what the Swedroe-type portfolio is all about, in my view. People don't want to make the tradeoff-- they are risk averse, but they are not willing to give up the "good right tail," the fantasy of getting rich, acquiring higher social status, and making others envious.
"My bond allocation is the amount of money that I cannot afford to lose." -- Taylor Larimore

michaelsieg
Posts: 553
Joined: Mon Jan 07, 2013 11:02 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by michaelsieg » Wed Jan 01, 2014 10:15 pm

Robert T wrote:
Michaelsig,

Here's the above analysis repeated with the AQR small cap momentum index in place of the MSCI momentum index. On its own, the AQR small momentum index almost matched the annualized returns of the S&P500 from 1984-1999. So little tracking error over this period. And 2000-2012 performance was higher, due to the smaller cap tilt of the AQR fund. The combination with RAFI pure SV, doesn't work as well as the MSCI momentum index, but there still seems to be some benefit - lower 2000-2002 downside, not as great a tracking error if using a small value fund only in 1984-1999, and higher returns in 2000-2012. Looks okay to me. The diversification benefit seems to come from a paring of momentum with value, so perhaps in combination with a value fund (of any market cap) there could be some value added. Obviously no guarantees.
Robert, I somehow missed your answer before and just looked at it now - thank you so much, this is very helpful.
Happy New Year! - I will try to focus more on sport prices (and less on spot prices...)
Michael

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Rodc » Thu Jan 02, 2014 9:07 am

Only "fools" confuse strategy and outcome.


While absolutely true it raises a couple of issues. Fifteen years is a near eternity when you are living it, if surprisingly brief when you look back at it. I think it would be extremely difficult for the vast majority of real life humans to watch such under performance and stick to an extreme tilt. And sometimes you really do have a poor strategy and it really is born out in the data. If you have done something stupid you don't want to keep at it for an entire lifetime. How is one supposed to tell the different between (1) good strategy that has done poorly for a decade or two but should be expected to someday turn around and (2) a poor strategy that has done poorly for a decade or two and should not be expected to turn around ever? How do you know you did not just get sucked into a good sounding, but ultimately not sound approach?

A counter to your bromide is "Only the "insane" keep doing what does not work expecting a different outcome next time."

Personally I think the tilt approach has merit. I have a somewhat tilted portfolio: about half way between the extreme contemplated in this thread and pure cap weight. The hope is that tilting for all its potential long underperformance really is a long term sound approach and I'll get about half of my cake and get to eat it too, but if I am fundamentally stupid, I won't have shot my self in the foot too badly. Of course the bonds will mute things so I don't expect any miracles of outperformance: any benefit or harm is expected to be modest.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

lazyday
Posts: 3300
Joined: Wed Mar 14, 2007 10:27 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by lazyday » Thu Jan 02, 2014 5:06 pm

Rodc wrote:(snip) How is one supposed to tell the different between (1) good strategy that has done poorly for a decade or two but should be expected to someday turn around and (2) a poor strategy that has done poorly for a decade or two and should not be expected to turn around ever?
Also over 15 years, there are often large changes in academic research results, academic opinion, and guru opinion.
And available investment choices, such as asset classes.

larryswedroe
Posts: 15726
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by larryswedroe » Thu Jan 02, 2014 8:00 pm

Rodc
Your observation about people's ability to stay the course during periods of underperformance is exactly why it will always IMO carry a big premium--most investors cannot emotionally stay the course--it's also why they fail as investors with great persistence.
The success of any strategy IMO depends the most on whether you as the investor believes that the strategy is right and will thus NEVER confuse strategy and outcome--in other words a strategy must be right or wrong before you know the outcome
Larry

lazyday
Posts: 3300
Joined: Wed Mar 14, 2007 10:27 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by lazyday » Thu Jan 02, 2014 10:24 pm

larryswedroe wrote:The success of any strategy IMO depends the most on whether you as the investor believes that the strategy is right and will thus NEVER confuse strategy and outcome--in other words a strategy must be right or wrong before you know the outcome
Larry
I suspect that even many Bogleheads' beliefs weaken or change over time.

That said, I'm glad that you make posts that only some people might benefit from. Were I in your position, might be tempted to filter most of what I say, leaving little of any use.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Fri Jan 03, 2014 3:27 am

.
I think there are perhaps two seperate aspects being discussed.

1. A strategy itself.

2. Outcomes vs. strategy.

On the strategy itself

In my experience (outside of personal investing), the 'success' of a strategy is linked to how implementable it is. Some strategies look great on paper, but thats sometime were they remain. In my view implementation is integral to strategy - it should be considered in the formulation of the strategy itself to increase the likelihood that it will be implemented.
  • "Implementation [execution] is a dsicipline, and intergral [fundamental] to strategy, and has to shape it" from "Execution: The Discipline of Getting Things Done" by Bossidy and Charam
This is also a point made by Jack Welsh "In real life, strategy is actually very straightforward. You pick a general direction and implement like hell" - from "Winning". Although I believe that having focused targets (not general directions) helps to discipline implementation and associated action.

In this respect I think investors should consider implementation aspects in their investment strategy/plan (in this case the likelihood of sticking with it) - IMO this includes both likelihood, duration, and madnitude of downsides, and track error. Larry does cover this in his books.

On outcomes vs. strategy

I generally agree. The example on this often given is an investment manager going to Vegas betting all their clients money on red and winning. While increasing wealth dramatically overnight - a good outcome for the clients - its a poor strategy (IMO) - so we shouldn't confuse a good outcome with necessarily a good strategy and vice versa. But I do think we should have periodic reviews on how we are doing against targets, and what we are doing to get there.

Robert
.
Last edited by Robert T on Fri Jan 03, 2014 5:38 am, edited 1 time in total.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Fri Jan 03, 2014 3:55 am

.
Just to elaborate a little more on the earlier point on momentum. Here are some extracts from the Asness article in Value and Momentum Everywhere - Journal of Finance - June 2013

  • "Value strategies are postively correlated with other value strategies across otherwise unrelated markets, and momentum strategies are positively correlated with other momentum strategies globally. However, value and momentum are negatively correlated with each other within and across asset classes."

    "The negative correlation between value and momentum strategies and their high positive expected returns implies that a simple combination of the two is much closer to the efficient frontier than either strategies alone, and exhibits less variation across markets and over time."

This seems to suggest that adding momentum to a US value strategy is a more effect diversifier than adding International value re: the articles' findings that value strategies are positively correlated within and across markets, while value and momentum are not. If this is indeed the case, and given that the momentum premium has historically been large when the S&P performed well (1984-1999) adding momentum to a heavily small cap value tilted portfolio could be an effective way to reduce tracking error (adding as much as is needed to match expected 'tracking-error' tolerance) without giving up much in expected return (it seems much less so than if adding S&P500). The earlier posts provided some examples of this with available investment options - but with the caveat that this was demonstrated with backtested/not live data). Something to consider IMO. Just a caveat - the conclusions of the Asness paper are based on average correlations across markets within which there may be lagging countries e.g. Japan - so still think international diversification is important (i.e. adding momentum to a globally diversified portfolio, rather than as a substitute for global diversification).

Robert
Last edited by Robert T on Fri Jan 03, 2014 5:04 am, edited 1 time in total.

lazyday
Posts: 3300
Joined: Wed Mar 14, 2007 10:27 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by lazyday » Fri Jan 03, 2014 4:28 am

I suppose those without access to good ex-US SV funds could get some foreign currency exposure with unhedged bonds.

User avatar
Blue
Posts: 1151
Joined: Sat Jul 12, 2008 10:18 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Blue » Fri Jan 03, 2014 5:02 am

Perhaps a naive way of looking at this, but part of me sleeps better at night knowing that Bernanke/Greenspan/Yellen is interested and actively considering the outcome of beta risk. I'm not sure they have the same consideration for other factor risks. Trying to avoid being political, my concern would be that the Bernanke put only applies to beta risk. Ex ante, at least to me, it nudges consideration to minimize tracking error generically and specifically through adding beta rather than momentum to a tilted portfolio.

RNJ
Posts: 789
Joined: Mon Apr 08, 2013 9:06 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by RNJ » Fri Jan 03, 2014 5:32 pm

Robert T wrote:.
I think there are perhaps two seperate aspects being discussed.

1. A strategy itself.

2. Outcomes vs. strategy.

On the strategy itself

In my experience (outside of personal investing), the 'success' of a strategy is linked to how implementable it is. Some strategies look great on paper, but thats sometime were they remain. In my view implementation is integral to strategy - it should be considered in the formulation of the strategy itself to increase the likelihood that it will be implemented.
  • "Implementation [execution] is a dsicipline, and intergral [fundamental] to strategy, and has to shape it" from "Execution: The Discipline of Getting Things Done" by Bossidy and Charam
This is also a point made by Jack Welsh "In real life, strategy is actually very straightforward. You pick a general direction and implement like hell" - from "Winning". Although I believe that having focused targets (not general directions) helps to discipline implementation and associated action.

In this respect I think investors should consider implementation aspects in their investment strategy/plan (in this case the likelihood of sticking with it) - IMO this includes both likelihood, duration, and madnitude of downsides, and track error. Larry does cover this in his books.
. . .

Robert
.
Our capacity as human beings to analyze a strategy far outstrips our capacity to understand our ability to implement it. Tracking error is a finance issue. Tracking error regret is an emotional issue.

pradador
Posts: 156
Joined: Thu Jun 14, 2012 9:20 pm
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by pradador » Sat Jan 04, 2014 4:38 pm

Taking this to an extreme, I'm curious to see how this will behave into the future:

Code: Select all

7.5% PXSV - PowerShares Fundamental Pure Small Value Portfolio
7.5% PDP  - PowerShares DWA Momentum Portfolio
7.5% TLTE - FlexShares Morningstar Emerging Markets Factor Tilt Index Fund
7.5% PIE  - PowerShares DWA Emerging Markets Momentum Portfolio
35.0% SHY  - iShares 1-3 Year Treasury Bond ETF
35.0% VTIP - Vanguard Short-Term Inflation-Protected Securities ETF

lazyday
Posts: 3300
Joined: Wed Mar 14, 2007 10:27 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by lazyday » Sat Jan 04, 2014 8:41 pm

No need to split mom/SV equity 50/50. Lg momentum may have lower returns than deep SV, allowing less bonds, and momentum seems to have shakier theoretical justification. So may choose less than half of equity.

Also might be worth considering more TIPS than nominal Treasuries, if expected risk adjusted portfolio returns are higher.

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by grap0013 » Mon Jan 06, 2014 8:09 am

RNJ wrote: Our capacity as human beings to analyze a strategy far outstrips our capacity to understand our ability to implement it. Tracking error is a finance issue. Tracking error regret is an emotional issue.
+10 Well said.
There are no guarantees, only probabilities.

JohnnyFive
Posts: 157
Joined: Mon Sep 08, 2014 1:14 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by JohnnyFive » Thu Mar 19, 2015 9:39 pm

Robert, what sort of Alpha did you turn up in your analysis of the underlying index for MTUM?

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Maynard F. Speer » Fri Mar 20, 2015 8:20 am

I came to a similar conclusion managing my own money: That I'd never be comfortable with half my wealth in an index .. I often think the success of steadfast 60-40 investors is that it's quite a bold thing to do, and you do get rewarded for staying the course ..

Looking at Swensen's Yale portfolio, I found something more in line with my thinking - the kind of portfolio Efficient Frontiers models steer you towards - of offsetting higher risk investments (in this case private equity, factor tilts, emerging markets, and opportunistic positions) against more conservative capital preservers (Absolute Return and real assets)

I don't think the implementation matters a great deal, but the principle, for me, seems to create a more market-agnostic portfolio - and I have found momentum to be a useful diversifier too

Image
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Browser » Fri Mar 20, 2015 10:30 am

For fans of the Harry Browne Permanent Portfolio, it is noteworthy that over the 1984-1999 time period you would have ended up with a portfolio value that was just 20% of the value of the S&P 500 compounded annually. Comparing to the more conventional allocation of 60% stocks / 40% bonds you ended up with 33% of the value of the 60/40 portfolio. Talk about tracking error regret! How many of you think you could have held onto the PP over that period?
We don't know where we are, or where we're going -- but we're making good time.

Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Browser » Fri Mar 20, 2015 12:30 pm

I'm not clear Robert T. Do you hold any foreign stocks or are you all domestic? If you hold foreign stocks, how do you calibrate the tilt of these in your portfolio?
We don't know where we are, or where we're going -- but we're making good time.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Thu Mar 26, 2015 5:46 am

JohnnyFive wrote:Robert, what sort of Alpha did you turn up in your analysis of the underlying index for MTUM?
Zero - at least according to this earlier analysis viewtopic.php?p=2340505#p2340505
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Thu Mar 26, 2015 6:22 am

Browser wrote:I'm not clear Robert T. Do you hold any foreign stocks or are you all domestic? If you hold foreign stocks, how do you calibrate the tilt of these in your portfolio?
In equities: 50% foreign stocks. IMO no need to add intl momentum. US and Non-US developed momentum indices have, at least historically, been highly correlated, more so than value, and if I recall, slightly more so than market correlates. Wouldn't argue with those who add it, but don't think its really needed, particularly if not targeting a positive and significant momentum load at portfolio level.
.

pauliec84
Posts: 412
Joined: Wed Feb 09, 2011 11:29 am

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by pauliec84 » Thu Mar 26, 2015 9:42 am

JohnnyFive wrote:Robert, what sort of Alpha did you turn up in your analysis of the underlying index for MTUM?


Zero - at least according to this earlier analysis viewtopic.php?p=2340505#p2340505
Have you tried running the data since index inception (1975). Including data from the 70s, lowers alpha to a significant -0.24% monthly (~2.5% annual) when I run the analysis.
Coefficients
Intercept -0.024
Mkt-RF 1.042
SMB -0.212
HML -0.011
umd 0.321


Just to replicate 1980 onward
Coefficients
Intercept 0.001
Mkt-RF 1.05
SMB -0.20
HML 0.00
umd 0.32

And 1975 to 1980
Coefficients
Intercept -0.225
Mkt-RF 0.983
SMB -0.304
HML 0.090
umd 0.439

What the heck happened in the 70s. Maybe I have an error in my data.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Swedroe-type portfolio: downside risk vs tracking error

Post by Robert T » Thu Mar 26, 2015 4:21 pm

pauliec84 wrote:Have you tried running the data since index inception (1975). Including data from the 70s, lowers alpha to a significant -0.24% monthly (~2.5% annual) when I run the analysis
From 1975 monthly alpha on MSCI USA Momentum = -0.02, and on FF Large Cap Momentum = -0.07. Yes large negative alpha on both from 1975-79. My sense is this is random drift. I had earlier done a more detailed analysis of midcaps, and found a fairly signficant drift in alpha by decade - some flat, some positive, some negative. Over short periods the factor loads also tend to drift. My sense is this is simply due to the shifts in the drivers of the premiums themselves.

The premiums are comprised of performance differences across the segments that comprise the premiums (see links) - large, small, high, low. During different short-term time periods different segments likely drive the premiums more than other time periods (although perhaps more stable over longer periods) . If assessing Large cap momentum, or small value etc, the short-term alpha and factor load drifts are likely (IMO) due to short-term drifts in return differentials across the components of the premiums (e.g. large, small, high, low). Not perfect - but seems to call for relying on longer time periods than shorter periods to try to caputre long-term characteristics/factor loads/alpha of the underlying series.

http://mba.tuck.dartmouth.edu/pages/fac ... ctors.html
http://mba.tuck.dartmouth.edu/pages/fac ... actor.html

Robert
.

Post Reply