Many Seeking Alpha readers are dividend advocates so I put together a portfolio built around high yield ETFs. Readers can find this portfolio over at this link. I asked Ernie Stokely to back-test the performance and volatility of this dividend oriented portfolio to see how it compared to the “Rutherford 10” portfolio. Once more, the Rutherford 10 comes out on top. Why is the Rutherford 10 so difficult to top when it comes to performance, low volatility, and probability of beating the VTTVX benchmark?
If I recall correctly, the Rutherford 10 was put together using a combination of the Swensen Six and IVY 10 portfolios.
Dividend Correlated ETFs
Below are the correlation clusters of the dividend portfolio. As I look over this group of ETFs, many I counted on to produce return are highly correlated. I’m referring to these ETFs. [VTI, VYM, DVY, and IDV]
Rutherford 10 ETFs
For comparison, the correlation clusters for the Rutherford are shown below. From looking at these two sets of data it is not obvious why the Rutherford 10 performs better than the Dividend Portfolio. While there is overlap in a few ETFs, DBC and GLD, two low correlated ETFs, are missing from the Dividend Portfolio. Might those two missing ETFs make a difference?
Lowell,
Could it be simply that many of the dividend ETFs do not respond strongly to a bull market by their very nature of being filled with dividend-producing stocks, while Rutherford’s portfolio is laden with equity stocks that are the drivers of a bull market?
I doubt DBC and GLD will add much to the long-term performance because they are very seldom chosen by the momentum engine over the usual 10 year back test period.
Ernie
Ernie,
You may be right as to why the Dividend Portfolio lags the Rutherford 10. Even though VOE, VBR, and VOT are highly correlated with VTI, I’ve added them to a few portfolios in hopes of squeezing out a little more return.
TIP is an ETF that is part of the Rutherford 10 and I don’t recall it ever made the recommended list.
Lowell
Yes, TIP has been in the doldrums for a long time due to QE! Someday TIP is going to come out of the closet! Ernie
Lowell,
Here is an interesting post by Dorsey Wright Money Management: http://seekingalpha.com/article/4004856-upper-end?app=1&uprof=46&isDirectRoadblock=false This was published today. My take is that populating a momentum portfolio with ETF’s that have a good track record increases the odds for higher returns. I believe this probably applies to a sector rotation strategy possibly more so than a more broad based portfolio such as Rutherford but may have application there too. Hope this stimulates some discussion.
Richard
Richard,
Thank you for the link. I’ll defer to the back-testers such as Ernie on this one, but I think we found applying the momentum model to sectors did not work as well as the Rutherford 10. Maybe we need to test this again.
I don’t doubt that better performers should produce better momentum results.
Lowell
Richard,
Yes, this is essentially what we are doing in the “Kipling” momentum model (although we are not using Point & Figure charts for ranking – but the principles are similar). The results are consistent with our findings in that the conclusion is that we should use assets ranked in the top quintile – in the case of our 10 asset Rutherford portfolio this would be the top 2 ranked spots.
I don’t think it makes sense to directly compare the returns from a sector rotation model that is based on a shortlist of relatively highly correlated assets within a narrow asset class (US equities) to returns from the Rutherford that is a totally different portfolio composed of widely diversified “global” assets. However, the principles are the same in both cases – pick the top ranked assets in both cases.
The choice of assets to be included in the asset selection list has been widely discussed on this blog and comes down to an investor’s objectives. If my objective is to beat the performance of the broad US equity market then it might make sense to break that market down into it’s market sectors and invest in the best performing sectors. On the other hand, in times when US equity markets are not strong, this approach may only serve to minimize losses whereas a broader based portfolio might shift to Bonds or Commodities if these are highly ranked.
The other message is also quite interesting and re-enforces one of the problems I struggle with every time I face an adjustment – do I wait for a pullback before entering (since top ranked assets are usually near recent, and maybe all-time, highs) or buy/sell without too much attention to the chart. As I’ve mentioned in a few posts, sometimes (maybe always?) it’s better/easier/less frustrating to “Just Do It” – I think this is the message coming from Dorsey Wright.
Thanks for the link – the folks at Dorsey Wright are very good and worth listening to.
David
David, Lowell, Ernie,
David’s message to “Just Do It,” is supported by a study I read recently that opined that the reason the majority of actively managed investment accounts under preform indexes is because the managers don’t strictly adhere to their strategy. With that said, the market goes up much slower t
David, Lowell, Ernie,
Will try again. My last message disappeared of of my I pad screen and did not realize it posted. What I was trying to say is that the market goes up much slower than it goes down and being retired, I have found it much harder to keep calm and wait for the review date to make trades on a day like last Friday knowing there is a limited time for losses to be recovered. I usually regret it when I don’t wait.
Richard
Richard,
Discipline has a lot to do with successful investing. I tend to stick with my 33 day review schedule and pay close attention to the investment model. Judgment is also required. Like you, I am retired so I tend to tilt on the side of risk rather than return.
Investors in the ages 20 to 50 should tilt to the return side as they have many years to build a portfolio. The young should also be on their knees praying for the market to go down so they can buy more shares for the same amount of dollars.
Lowell
Richard,
Here is a portfolio construction model that is nearly impossible to back test unless done by hand.
1. I have a starting database of around 125 ETF that span the globe. This includes the 101 commission free ETFs from TD Ameritrade.
2. Seek the top performers from International Equities, U.S. Equities, Bonds, Dividend Driven, Commodities, and Precious Metals. We need to break the 125 into smaller groups as the Kipling will not handle more than 40 securities at a time.
3. Narrow the list to approximately 20 to 25 ETFs.
4. Run a correlation on the smaller list and then select one ETF from each cluster. This should narrow the list to 10 ETFs.
5. Build the portfolio from the top two or three ETFs that emerge from the cluster filter.
To back-test this portfolio construction model would take an enormous about of programming time and computer power.
Lowell
Lowell,
When you say, “Seek the top performers …” in 2., are you referring to their momentum scores, or …? Also, putting more than a dozen ETFs into the Kipling momentum engine I feel will not give one very reliable results.
I can put as many ETFs into my R back tester as my memory allows, but I’m not sure loading it with all these ETFs would give any useful results. I suppose one could “batch” them, and take the top 5 from each batch … But you are correct. It would be a big job.
Ernie
Ernie,
The ETF Ranking blog may explain what I had in mind. I began with some 125 ETFs from four major groupings. I used the Tranche Momentum spreadsheet for the screening process and ended up with 17 ETFs plus DBC for a low correlated ETF.
The final recommended ETFs were highly correlated – a problem in my opinion.
Lowell
Ernie,
I “batched” the ETFs into four groups.
Lowell
Lowell
When you run your correlations I take it that you are introducing risk parity based on past history which has, for the most part, been a period of lowering interest rates. The question is, how well will the correlations hold up during a period of rising interest rates which is the direction we seem to be headed for? Thought you would like an easy question today. 🙂
Richard
Richard,
Rising interest rates will likely impact VNQ and RWX, but I don’t know what the impact will be with the other asset classes. Correlations do not remain static. For example, it is not unusual to see VTI and VEA have a correlation of 0.80 or higher. That was not the case when I ran the above correlations.
Lowell
I think diversification becomes more obvious when we start to restrict the number of “clusters” (widen diversification ranges).
For example consider the above portfolios divided into 5 clusters :
The Dividend portfolio would consist of 5 clusters containing (VTI, VYM, DVY, HDV, JNK, HYG, IDV, PICK), REM, (VNQ, VOX), VPU and TLT (cutoff = 0.63 – 61% in 1st cluster).
The Rutherford Portfolio would break down into (VTI, VWO, VEA, RWX), DBC, (VNQ, PCY), (TIP, TLT), GLD (cutoff = 0.49, max – 40% in 1 st cluster).
I also think Ernie’s comment that the Rutherford is “laden with equity stocks” is a bit overstated – there are only 3 of the 10 ETFs in this asset class – and these do, in fact, cluster together as we might expect. Also, as I recall, DBC and GLD did have their moments in the sun around the time of the 2008 equity market pullback – exactly what we want. In the end we want the best total returns (growth plus dividends) and we rely on the momentum engine, using adjusted prices, to detect these assets. And yes, we may be glad to have TIP in our lists when inflation returns.
Bottom line, as we all know, performance is strongly influenced by the assets we have in our lists – it’s just a matter of which ETFs we choose to include.
David
Lowell,
You might want to try adding DBC and GLD to the Dividend portfolio and break the portfolio into five clusters to see how this splits the list. This may answer Ernie’s question about the significance of these asset classes.
David