Magic Formula on the ASX

[systematic]

As a change from the tiny, illiquid, unprofitable companies in the NCAV thread, but still in keeping with the theme of value based investing, I thought I'd take a look at the magic formula on the ASX. This is the well known method from, "The Little Book that Beats the Market" by Joel Greenblatt. There's a brief wikipedia article with the details, if you are not familiar.

I post the top 10 rather than the top 30 or 50 as Greenblatt really sticks to the top few percentiles - and the ASX is a much smaller universe of stocks to select from.

Currently, I get:



What's your analysis of any of the stocks in this list?

My initial comment: at least you don't just get a bunch of mining stocks (as you do with the NCAV and some other singular value measures). For interest, the next 5 contain both Nine Entertainment and Southern Cross, so there's at least a little bit of an emphasis on media stocks at the moment.

 

[sinner]

Thereis an excellent bunch of writing about the Magic Formula in "Quantitative Value", I recommend reading it, if you are relying on the Magic Formula!

The main issue with this methodology is that it ranks value and quality equally, which generally results in portfolios that overpay for quality.

A better solution (and that adopted by the authors of Quantitative Value) is to split the universe into deciles based on EBIT/TEV, select the value decile and then split your selection into high quality and low quality halves.

The returns on High Quality Value are better than Low Quality Value, Value and Magic Formula.

Here are all the analysis on Magic Formula done by the QVAL guys: http://blog.alphaarchitect.com/tag/magic-formula/

The book also has some really interesting observations (not necessarily negative) about this methodology and the general issue of mistaking the ceiling of a methodology for its floor.

 

[systematic]

Thereis an excellent bunch of writing about the Magic Formula in "Quantitative Value", I recommend reading it, if you are relying on the Magic Formula!

The main issue with this methodology is that it ranks value and quality equally, which generally results in portfolios that overpay for quality.

A better solution (and that adopted by the authors of Quantitative Value) is to split the universe into deciles based on EBIT/TEV, select the value decile and then split your selection into high quality and low quality halves.

The returns on High Quality Value are better than Low Quality Value, Value and Magic Formula.

Here are all the analysis on Magic Formula done by the QVAL guys: http://blog.alphaarchitect.com/tag/magic-formula/

Thanks sinner, I'm familiar with that work, and prior to the QV book coming out had concluded that you are better off with value first, then quality, as opposed to an equal split (you could see that from looking at Novy-Marx' data, which one of the co-authors of QV mentioned at the time as well). That's if you're using quality at all. I've previously mentioned that I'm quietly cautious on quality for several reasons - one of the co-authors of the QV book doesn't like to use it at all.

I was crook as a dog the other day and grabbed "The Little Book that Beats the Market" off the shelf for a quick read in bed with a cuppa (Greenblatt has a very entertaining writing style)...and that's what reminded me of this approach.
They (NCAV, MF, QV etc) are not my investment plan, as it were - I just find looking at stuff to be fun.

In my view, at the end of the day they are slightly different approaches that will produce slightly different numbers. Most of these things are variations on a theme, to various degrees. Any quality measure combined with any value measure (equal weighted) gives you an MF type of approach (e.g. Price to Book with Gross Profits to Assets). Or, for example, the Piotroski approach written about 13 years ago (quintile on P/B first, then a top F Score) is the exact same theme going on with QV. And, in fact, compares quite favourably with the more complex QV.

I've had different views (as we all do, as we keep thinking about these things). It all depends on what the investor is trying to achieve. For example: some investors might like to start with 'quality' (however they define it). They might find a quantitative sort on quality measures to be an initial, very useful cull of the market...to then go on and do their own intrinsic value calcs on. Certainly not my approach - but entirely valid. Greenblatt talks about this in relation to the Magic Formula (and apparently it's what he actually did / does in his funds). It depends what the investor is trying to achieve (which of course, is not always the highest CAGR, as much as we'd like it).

Anyway, it's all food for thought and I really appreciate the input. I also completely agree with you that QV is a wonderful summary of - just that: a quantitative approach to value, and I regard the work extremely highly, as you obviously do. As a matter of fact, I started putting it together on the ASX a couple years back and then abandoned it. Might have to look at it again sometime.

 

[sinner]

Awesome post, systematic, it's obvious my cautionary note was not required for you!

I also noticed how well the Piotroski/FS_SCORE + P/B value decile performed (I posted this link in another thread recently but it's even more relevant here) http://blog.alphaarchitect.com/2015...ple-methods-to-improve-the-piotroski-f-score/ with CAGR and maxDD approaching that of QV for a lot less effort.

Thanks for reminding me about this as I had some research to do on P/B based on this regression model which works off P/B and RoE, which for me was a completely different way of looking at the same data http://epchan.blogspot.com.au/2014/02/fundamental-factors-revisited-with.html .

I had also considered setting up QVAL portfolio for the ASX, same as you, but in the end I didn't mostly because there is something about the logits used in the book which rubs me the wrong way (same reason I don't like Altman z score). So in the end I adopted a much simpler model (read, robust) influenced by the work of Eric Falkensteins DefProb (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1103404) to replace the logits and use that as part of my ASX portfolio selection.

 

[ThirtysixD]

Nice to see some people with similar thoughts!

Most of my research has shown that roe outperforms roic (roce also beats roic for that matter). I think quality tends to mean revert so filtering according to value is more important. Instead of ranking on quality using a straight roe > 0 filter will work just as well if not better. However this can be problematic on the asx since we are not very well diversified.

I am currently holding 3 of your stocks from your list - but I tend to focus on small caps on the asx (less efficiently priced and room to grow). I think there are several stocks that would rank higher than those listed so im curious about where you are getting your data from?

What I have been thinking about recently is front running these quantitative investing funds. Most of their strategies are fairly simple and if they were to ever get a decent amount of assets under management you would only need to figure out the holding period to pick them off.

 

[sinner]

What I have been thinking about recently is front running these quantitative investing funds. Most of their strategies are fairly simple and if they were to ever get a decent amount of assets under management you would only need to figure out the holding period to pick them off.

I'm not sure what the point of that is exactly, feel free to enlighten me!

I'm an investor because it's the optimal method for me, all the turnover is handled inside the funds! But if anything, I wish they had a longer holding period, more like allocate 20% of capital each to 5 portfolios (generating a new one and dropping the oldest each year) with 5 year holding period.

 

[DeepState]

Given mention of Novy-Marx, various profitability measures, Piotrosky and interaction with value....

Might be worth reviewing this:

http://www.researchaffiliates.com/O...s/259_The_Moneyball_of_Quality_Investing.aspx


Of note, these guys reckon there is no premium to quality in a univariate sense. Further, the quality measures which are presently popular are likely data snooped. However, they seem to think that conditioning value by quality produces good stuff. ... Which suggests there is mispricing of value stocks that can be picked up via quality metrics.



Asness finds that Quality Minus Junk does produce a premium.

Arnott (of RAFI: Research Affiliates Fundamental Index per above) and Asness used to be good buddies, but not any more.

This thread is fantastic. It is a treat to find highly data rational people coalescing.

Question: Do you think that Quant managers and those with the simple resources to replicate these measures in a day or two, who read this research and more, will not be front running you as you try to front run them? I express some skepticism when there is an argument based on non-linear interaction and mispricing is asserted.

 

[tinhat]

Currently, I get:

View attachment 63312

What's your analysis of any of the stocks in this list?

Most of those stocks are in a serious down trend. This may indicate the problem with associating a stock's market price (which is the market's forward looking assessment of the company) with a company's past performance, which is what equity yield does.

 

[systematic]

Most of those stocks are in a serious down trend. This may indicate the problem with associating a stock's market price (which is the market's forward looking assessment of the company) with a company's past performance, which is what equity yield does.

a) I doubt it. It's only been two months (the MF holding period is one year). Regardless, the portfolio is down 4.5% whilst the all ords is down 8%.

b) The MF uses an earnings yield, not an equity yield just out of interest. But to your point re: a company's past performance...what else is there? No one has a company's future performance, so the only other alternative would be to be, "forward looking" (as you put it) with the company's performance. Well that's just another topic altogether, and a dangerous road to travel, in my opinion. That domain is for super investors, not people like me. Of course, some pull it off (Buffett, ASF's own, craft). I can't punt on myself being that good.

 

[systematic]

Of note, these guys reckon there is no premium to quality in a univariate sense. Further, the quality measures which are presently popular are likely data snooped. However, they seem to think that conditioning value by quality produces good stuff. ... Which suggests there is mispricing of value stocks that can be picked up via quality metrics.

I agree, there's so much data snooping going on with the latest and greatest factors. My take on quality thus far has been to ignore it (due to similar sentiment as above). The only use for quality that I see is in picking up genuine bargains, or mispricings, as you mention (value first, then quality). It just makes sense, there's data to show it, and I'm all for reducing volatility a little. So, I've been looking at it a bit...but even then I still remain cautious - as I lean a little more toward the school of thought that says: any factor you decide to include in your model, should be a stand-alone factor. Actually, Novy-Marx just recently had a paper on exactly that. I'm still definitely undecided at the moment whether to include it, so it sits on the sidelines for now - although I look at it a lot.

 

[sinner]

Given mention of Novy-Marx, various profitability measures, Piotrosky and interaction with value....

Might be worth reviewing this:

http://www.researchaffiliates.com/O...s/259_The_Moneyball_of_Quality_Investing.aspx


Of note, these guys reckon there is no premium to quality in a univariate sense. Further, the quality measures which are presently popular are likely data snooped. However, they seem to think that conditioning value by quality produces good stuff. ... Which suggests there is mispricing of value stocks that can be picked up via quality metrics.

The main goal for me when applying quality measures is to avoid value traps.

I agree about the quality issues you mentioned re snooping and lack of premium. My main concern is the introduction of a variable highly influenced by cyclical factors. If you read the QV book they have a decent (although IMHO not perfect) method to address this, they take the 8 year geometric average of ROA (example). The geometric average and 8 year both underweight the cyclical portion.

Re lack of premium, I think it is important to recognise (a realisation I am only just coming to), for example RoE based portfolio formation might not produce a portfolio of outperforming stocks but those stocks are actually less likely to experience significant drawdown relative to the benchmark (i.e. high correlation with low beta)! Also interesting to note, if you look at the epchan link I posted, that there is starting to find evidence that while "quality" doesn't necessarily influence returns systematically across a universe of stocks, it does influence the future returns of single names along with value metrics. My guess is basically this is capturing the business/profit cycle in a way that value measures never can.

Asness finds that Quality Minus Junk does produce a premium.

Others too, but how they define this kind of annoys me. If you look at the equity curve you can see it is basically long volatility. You can produce a nearly identical curve by plotting the ratio of SPLV to SPHB ETFs (100 lowest histvol SP500 stocks vs 100 highest beta SP500 stocks rebal annually) or DEF vs SPY or whatever similar and all of the returns come from drawdown periods in the market. This is very similar in concept to how the returns of low momentum stocks are actually positive most of the time. As I said, this annoys me because better returns can be achieved without the silliness of defining things as a premium. i.e. may as well long high momentum (ignore short low momentum) during bulls and short high beta (ignore long low vol or high qual) during bears. The return premium identified comes from the market regime, not from the stocks themselves, per se, the alpha comes from the managers ability to identify and time the market as usual.

Question: Do you think that Quant managers and those with the simple resources to replicate these measures in a day or two, who read this research and more, will not be front running you as you try to front run them? I express some skepticism when there is an argument based on non-linear interaction and mispricing is asserted.

Agreed 100%.

 

[tinhat]

a) I doubt it. It's only been two months (the MF holding period is one year). Regardless, the portfolio is down 4.5% whilst the all ords is down 8%.

b) The MF uses an earnings yield, not an equity yield just out of interest. But to your point re: a company's past performance...what else is there? No one has a company's future performance, so the only other alternative would be to be, "forward looking" (as you put it) with the company's performance. Well that's just another topic altogether, and a dangerous road to travel, in my opinion. That domain is for super investors, not people like me. Of course, some pull it off (Buffett, ASF's own, craft). I can't punt on myself being that good.

*earnings* yield. my bad.

I'm a FA investor and weighted towards income (dividend) with the aim of holding long term, and I don't use a precise model, which is what you are looking for, but, I do look at a couple of things:

- EPS growth. Have earnings over the past few half yearly reports been growing (backward looking but trend seeking)?

- Forecast EPS growth. For what they are worth, consensus targets as provided by Thompson Reuters or Morningstar. Is the positive EPS trend forecast to continue? Where there are less than five analysts making up the consensus forecast, the fewer the number of analysts in the forecast the less weight you give to the forecast.

Personally, if i was using the magic formula model (and I had never hear of it until this thread), I would consider running the results of the earnings yield filter through something that filters for earnings growth trend at least. But I am not a t/a investor, and I don't even know how to back test or run monte carlo simulations, etc.

 

[systematic]

That's all cool, tinhat. Differences are what make the market!

My only thought to add is that everything I've read has put me off looking at growth - particular longer term growth. A little bit of recent growth is okay.

Re: using forecasts for trend (as opposed to accurate forecasts) - I think that's fair enough. Even David Dreman thought that was okay. I just don't think it's necessary to use. One example summarised from the studies I've read: which performs better, a P/E using trailing earnings or a P/E using forecast earnings (or even a 50/50 blend?). The trailing earnings win. Forecasting earnings trends and looking for past growth trends misses out on mean reversion of earnings. That's just my take, anyway. I like that we all come at this differently!

 

[ThirtysixD]

I'm not sure what the point of that is exactly, feel free to enlighten me!

I'm an investor because it's the optimal method for me, all the turnover is handled inside the funds! But if anything, I wish they had a longer holding period, more like allocate 20% of capital each to 5 portfolios (generating a new one and dropping the oldest each year) with 5 year holding period.

Im getting a bit off topic here but even as a mid-late comer you can get an extra 1-2% pa from front running the Russell 2000. 100 or 200 million divided by 50 stocks is really nothing but once you start talking higher aum then etfs can start moving stock prices. Not saying this will be the latest holy grail craze but its something that I have always found interesting.

IDOG for example has 1 bill aum and purchases 50 stock according to yield annually. This leads to 20 mill buying pressure within a short period of time. Might not seem like much for big caps but there are less liquid stocks where this figure is very close to the daily average trading volume.

 

[ThirtysixD]

My mistake above... the us domestic fund is SDOG not IDOG

 

[DeepState]

​

My mistake above... the us domestic fund is SDOG not IDOG

This front running idea is a nice thought and an evolution of the index adds/deletes trade. Managers respond, though. This from the prospectus of the SDOG:

"The Fund may sell securities that are represented in the
Underlying Index or purchase securities that are not yet
represented in the Underlying Index in anticipation of their
removal from or addition to the Underlying Index."

They will game the gamers. Also, this is an SP500 universe, so none of them is anything other than a liquid monster, even heading into Christmas period when the index is reconstituted.

It is a good thought though. The more obvious the rule, the easier it is to simply replicate, the more that trading it will impact prices....the more counter-measures will be in place. The juice lies with front running the not so obvious quant strategies....


Vanguard has over USD 1bn tracking the Russell 2000 index. Add BlackRock, State Street... and you have some real money.

 

[tinhat]

That's all cool, tinhat. Differences are what make the market!

My only thought to add is that everything I've read has put me off looking at growth - particular longer term growth. A little bit of recent growth is okay.

Re: using forecasts for trend (as opposed to accurate forecasts) - I think that's fair enough. Even David Dreman thought that was okay. I just don't think it's necessary to use. One example summarised from the studies I've read: which performs better, a P/E using trailing earnings or a P/E using forecast earnings (or even a 50/50 blend?). The trailing earnings win. Forecasting earnings trends and looking for past growth trends misses out on mean reversion of earnings. That's just my take, anyway. I like that we all come at this differently!

No doubt about it, I don't have the time or inclination to be a trader. There is a wealth in this world to be experienced that far exceeds the value of money. I am however, fascinated by those traders and T/A proponents on these boards that are willing to explain in simple words their ideas and TA ideas and open them up for consideration. A bit of both I say.

I have a very dear and very senior (in age) mentor (an incredibly patient and forgiving friend). My conversations with him are like being interrogated by Philip Fisher. He asks me about the management of any firm I invest in as if he might know the CEO's grandfather.

 

[systematic]

Most of those stocks are in a serious down trend. This may indicate the problem with associating a stock's market price (which is the market's forward looking assessment of the company) with a company's past performance, which is what equity yield does.

a) I doubt it. It's only been two months (the MF holding period is one year). Regardless, the portfolio is down 4.5% whilst the all ords is down 8%.

...A meaningless update at 15 weeks. Portfolio up 5.5% (from 6 losers, 4 winners). All Ords down -3%

 

[systematic]

^ Well, that was a bit of a premature evaluation. Today's move on (particularly) DLS and ACR makes the portfolio 10.8% against all ords -1.5%

 

[Ausinvest]

Hi Systematic,

Thanks for posting this. I'm new to this whole game and have been doing a little bit of reading on the magic formula.

Out of curiosity, can you tell me what criteria you used to get the 10 stocks above? Did you use ROA and P/E as per the little books instruction? What did you set your market cap, ROA and p/e minimums at?

Running my own numbers at the moment, if I take the books instructions word for word (I.e. ROA > 25% and P/E > 5, excluding financials and utilities), I get a pretty small list!

I'm still reading and learning, with a while to go before I actually throw some money into anything, but I'd really appreciate hearing how you got your numbers. Also - can you tell me what your list would be if you re-ran it today - if not too time consuming?

Many thanks