Sunday, January 04, 2015

MFI Select vs MFI Formula

I was asked about the difference between my MFI Select tranches and my MFI Formula tranches. I thought it might be easier to write a quick post.

MFI Select is what I have been doing since August 2012 when I re-booted my approach to MFI.  It consists of 5 stocks selected by me once a quarter. The stocks HAVE to be held for a year.  I select the stocks based upon my regular top 200 screens (MFI Top 200 List - 12/28/14). So it is possible I will get some stocks that are not from the official screen.  My selections are judgmental.  Here is how all the tranches have fared back to August 2012:

8/15/12 Stocks Start Current Dividend Pct Gain R3K Gain
DLB $34.43 $33.04 $4.00 7.6% 22.1%
KFY $13.98 $18.95 $0.00 35.6% 22.1%
MSFT $30.19 $31.92 $0.92 8.8% 22.1%
NSU $3.62 $3.33 $0.12 -4.6% 22.1%
UIS $19.96 $24.52 $0.00 22.8% 22.1%
Totals 14.1% 22.1%
11/15/12 Stocks Start Current Dividend Pct Gain R3K Gain
AAPL $537.62 $527.28 $8.75 -0.3% 36.7%
ABC $40.21 $69.23 $0.86 74.3% 36.7%
LPS $23.89 $34.26 $0.40 45.1% 36.7%
VIAB $49.12 $80.94 $1.15 67.1% 36.7%
WU $12.77 $16.40 $0.50 32.3% 36.7%
Totals 43.7% 36.7%
2/1/13 Stocks Start Current Dividend Pct Gain R3K Gain
AAPL $450.60 $500.66 $8.75 13.1% 21.1%
GA $6.10 $11.01 $0.65 91.2% 21.1%
STRZA $16.06 $27.98 $0.00 74.2% 21.1%
CA $25.07 $32.08 $1.00 32.0% 21.1%
NUS $41.94 $116.59 $1.20 180.8% 21.1%
Totals 78.3% 21.1%
5/6/2013 Start Current Dividend Pct Gain R3K Gain
KLIC $11.40 $14.44 $0.00 26.7% 18.9%
CYOU $29.24 $28.02 $0.00 -4.2% 18.9%
RDA $9.74 $16.79 $0.10 73.5% 18.9%
NVDA $13.85 $18.43 $0.32 35.4% 18.9%
CF $183.97 $243.22 $2.80 33.7% 18.9%
Totals 33.0% 18.9%
8/15/2013 Start Current Dividend Pct Gain R3K Gain
CF $187.26 $247.43 $4.50 34.5% 19.4%
FLR $66.06 $73.03 $0.53 11.4% 19.4%
AGX $15.60 $36.71 $0.75 140.1% 19.4%
KLIC $11.29 $14.26 $0.00 26.3% 19.4%
RPXC $15.92 $15.09 $0.00 -5.2% 19.4%
Totals 41.4% 19.4%
11/15/13 Stocks Start Current Dividend Pct Gain R3K Gain
AVG $17.22 $19.21 $0.00 11.6% 14.5%
ATVI $17.71 $20.10 $0.20 14.6% 14.5%
GA $9.04 $11.90 $0.23 34.2% 14.5%
RPXC $17.04 $13.91 $0.00 -18.4% 14.5%
SYNA $48.01 $61.90 $0.00 28.9% 14.5%
Totals 14.2% 14.5%
2/1/14 Stocks Start Current Dividend Pct Gain R3K Gain
ATVI $17.13 $20.13 $0.20 18.7% 15.9%
KLAC $61.47 $70.29 $18.40 44.3% 15.9%
RPXC $16.22 $13.71 $0.00 -15.5% 15.9%
SNDK $69.55 $97.88 $0.83 41.9% 15.9%
WNR $39.11 $38.36 $3.04 5.9% 15.9%
Totals 19.1% 15.9%
5/6/2014 Start Current Dividend Pct Gain R3K Gain
AVG $19.57 $19.58 $0.00 0.0% 10.3%
CA $29.54 $30.69 $0.75 6.4% 10.3%
CSCO $22.79 $27.61 $0.38 22.8% 10.3%
NSR $25.82 $27.74 $0.00 7.5% 10.3%
RCII $28.78 $35.95 $0.70 27.3% 10.3%
Totals 12.8% 10.3%
8/15/2014 Start Current Dividend Pct Gain R3K Gain
CSCO $24.40 $27.61 $0.19 13.9% 5.6%
AGX $36.71 $33.52 $0.70 -6.8% 5.6%
ITRN $21.32 $21.69 $0.38 3.5% 5.6%
BRCD $9.34 $11.72 $0.04 25.9% 5.6%
CBI $58.25 $42.05 $0.14 -27.6% 5.6%
Totals 1.8% 5.6%
11/15/2014 Start Current Dividend Pct Gain R3K Gain
FLR $64.80 $60.69 $0.21 -6.0% 1.3%
IPCM $38.45 $45.75 $0.00 19.0% 1.3%
VIAB $72.29 $75.40 $0.33 4.8% 1.3%
NHTC $12.19 $11.33 $0.01 -7.0% 1.3%
AWRE $4.40 $4.50 $0.00 2.3% 1.3%
Totals 2.6% 1.3%
Category/Tranche August November February May Total
Initial Investment      25,000       25,000     25,000      25,000    100,000
Current Tranche 1.8% 2.6% 19.1% 12.8% 8.8%
Previous Tranche 41.4% 14.2% 78.3% 33.0% 39.5%
Tranche -2 14.1% 43.7%            -               -   14.5%
MFI Overall Gain 64.2% 68.4% 112.2% 50.1% 73.7%
Current Balance      41,053       42,106     53,053      37,518    173,730
R3K Current Tranche 5.6% 1.3% 15.9% 10.3% 8.3%
R3K Overall Gain 54.1% 58.5% 40.4% 31.1% 46.0%
R3K Balance      38,514       39,622     35,102      32,783    146,022
Annualized IRR 23.1% 27.7% 47.9% 27.6% 31.5%

So as you can see, they have done very well. Up 73.7% in aggregate in the 2+ years.  If I had invested my money in 4 equal installments in Russell 3000 index, I would be up 46.0%.

MFI Formula Approach

I have noted in my analysis of all my MFI tracking portfolios (top 50 stocks > 100m market cap going back to 2006) that dividend stocks (which I have defined as a yield of 2.6% or greater) have way out-performed the broader class of all MFI stocks (December 31st, 2013 Monthly Tracking Portfolio).  And if you dig even further, larger market cap dividend stocks have done even better.

So starting September 30th, I have created a purely formula approach to MFI that picks 5 stocks greater than 2.4% yield and $600m market cap from the official screen.  This approach uses a random number generator and is described here (One Day To Next Formula Tranche).  I use the randomizer to take my bias out of the approach.

While this approach is new in my actual portfolios, I have back tested it. Here are my actual results:

10/1/2014 Stocks Start Current Dividend Pct Gain R3K Gain
BAH $23.52 $26.56 $0.11 13.4% 5.1%
BKE $45.55 $52.73 $0.22 16.3% 5.1%
CA $27.87 $30.69 $0.25 11.0% 5.1%
GME $41.70 $33.80 $0.33 -18.2% 5.1%
NUS $44.56 $42.24 $0.35 -4.4% 5.1%
Totals 3.6% 5.1%
12/31/14 Stocks Start Current Dividend Pct Gain R3K Gain
CA $30.83 $30.69 $0.00 -0.4% -1.2%
CSCO $28.36 $27.61 $0.00 -2.7% -1.2%
RGR $34.81 $34.90 $0.00 0.2% -1.2%
GME $34.21 $33.80 $0.00 -1.2% -1.2%
IQNT $19.87 $19.13 $0.00 -3.7% -1.2%
Totals -1.6% -1.2%

Here is the broad back test, not subdividing for $600m+ market cap.

Category Value
Total     186,864
Total Russell 3K     192,303
New     190,631
Dividend     430,270
So 100,000 spread over 12 monthly portfolios back in 2006 would be worth $186,864 just going with 50 MFI stocks each month from official screen. But would be $430,270 going with the dividend subset. I don't need to highlight that that is a material difference.  And if you went with large caps (over $600m), you'd be even higher.  The results were so good, I have actually pulled them from my postings - as I have decided to keep it proprietary.

Good luck everyone in 2015!

14 comments:

Unknown said...

Hi Marsh,

thanks for your update. I would like to start using the MFI method with the Paris Stocks Market, but I struggle to find databases with EY and ROIC of French companies I can test.
Do you have any hints apart from subscribing to ValueStockScreener?

Thanks in advance and wish you the best!!

Unknown said...

Marsh,

I agree totally that this needs to be proprietary if it has a chance to work.
If it were me it would be family only.

Jim Bowerman said...

is there a risk your just looking at a sector that has outperformed for a short while since 2007? dividend stocks in general have been on a tear since 2007. But will it continue? im not so sure

see this graph on dividend valuations:

http://facevaluewealth.com/wp-content/uploads/2013/10/dividend-stocks-are-expensive-2013.jpg

CorpRaider said...

It seems to me that it could be that (the quantitative easing/general affinity for yielders) or it could be the market cap and yield function as additional quality screens. DVP is an ETF that seems to be implementing a somewhat similar strategy. Top 10 cheapest S&P 500 stocks, weighted by cheapness (based on TEV/EBITDA) and screened for positive operating CF, dividends and b/s strength. I am also monitoring QVAL, which is basically a tweak of the MF as discussed in Gray and Carlisle's book Quantitative Value.

Unknown said...

Marsh, could you provide some info on the backtesting strategy you used? I tried many different combinations in Portfolio123, but in none of the scenarios did the dividend make a big difference in the time frame you mention (from 2006). In fact, in many scenarios, it underperformed a portfolio which did not have the dividend rule. Perhaps the way I am testing is incorrect.

jb said...

I'm having roughly similar thoughts as Corpraider, perhaps the backtesting performance is coincidental to other big economic trends over the past several years. The new/revised strategy reminds me of The Dogs of the Dow, but using a different and larger pool to pick selections from. If it works (and it should if you buy in to the MFI strategy), I say go for it. You've already narrowed down the stock universe to a set of stocks that should on average outperform the market.
cheers,
j

Marsh_Gerda said...

Great questions and comments. I have wondered also whether the broader out performance of dividend stocks since 2006 (makes sense with tax law changes and low treasury yields) has caused that component of MFI to do better than the rest. I really have no great way to answer that challenge. That is one reason I am only going 50% into MFI formula, the other 50% is still in select. Of course I have been thinking those points for past couple of years and missed out on some pretty snappy moves by MFI dividend stocks.

Marsh_Gerda said...

Andy - I have actually not done any back testing. I have simply been maintaining my own database going back to January 2006.

CorpRaider said...

Yeah, well you're fishing in a good pond as Mohnish Pabrai might say by selecting from amongst the MF stocks and then you're keying on a signal that "yeah we're less likely to be a Chinese internet fraud because we actually pay out cash." So you could be onto something. Meb Faber has published a little e-book about shareholder yield (roughly dividends and buybacks) I don't see why starting from MF stocks would do anything other than increase the probability of enhancing that.

Unknown said...

Marsh, thanks for the response. I did some rolling backtests for the timeframe 2006-now and also for the past 15 year time frame. From over 400 backtests, in the longer timeframe dividend rule underperforms normal MFI by about 3 percent, while in the 2006-now timeframe, dividend rule outperforms by about 2 percent. This seems very similar to the recent underperformance of small caps compared to large caps.

I think you are doing the right thing by not going 100% in the formula version.

Unknown said...

After digging a bit more, I think I know one explanation for this behavior: in the recent years, large caps have outperformed small and mid caps (http://www.dusthimer.net/Magic-Formula-Data.html). When you use regular MFI and just use a market cap of 600M, you end up with a lot of mid-caps in your portfolio. However, when you impose the additional dividend rule, the amount of mid-caps become very low and more large-caps end up in your portfolio.

To test this hypothesis, I re-ran the backtest for recent 5-10 years with regular MFI but with a market cap of 2b. This produces nearly same returns as 600m with dividend rule.

There might be more to this, but this may be one reason.

Unknown said...

The above comments don't take into account that anybody looking for Dividend income and doing a little research is not going to buy NUS, GME, RGR, COH, KING, and probably not even IBM.

These are not stocks that appear to be outperforming as they drop in value most every week.

If this works it because they are selling at bargain prices and nobody will touch them.

CorpRaider said...

Well no opinion on my part, but I know IBM is in VIG's top 10 holdings and the other ETFs that track the dividend achievers index. Not sure if they are in DVY, but they're almost certainly in numerous Wisdom Tree indexes that are weighted on dividends. Of course no outperformance likely came from IBM and COH over any recent period. Truly dogs with fleas. I'm long COH. hah.

Unknown said...

Sorry to pop up again, but do any of you have tips to gather data for backtesting Magic Formula in the European markets ?

I have subscribed to ValueStockScreeener, paying a monthly fee, but no historical data are available in their website