Share this page:
Stay connected:
Welcome to the Citywire Money Forums, where members share investment ideas and discuss everything to do with their money.

You'll need to log in or set up an account to start new discussions or reply to existing ones. See you inside!

Notification

Icon
Error

What do you take "Quality" companies to mean?
Tim D
Posted: 20 May 2018 21:37:21(UTC)
#21

Joined: 07/06/2017(UTC)
Posts: 447

Thanks: 1638 times
Was thanked: 669 time(s) in 294 post(s)
Good points.

Looking at the chart of IWFQ for its short life so far, I do have to wonder a bit why I'm bothering with it rather than just adding to my existing pure market-cap-weighted index holdings ...

Quality vs Vanilla

What it came down to was a desire to - with this year's fresh cash - invest it more in Microsoft and Unilever and Diageo type things. IWFQ's index's approach seems to result in a reasonably plausible distortion of the parent index, so I took a punt. (Although looking at the big companies which are conspicuous by their absence in IFWQ - e.g Facebook/Alphabet/Amazon/Intel, I suspect I could have achieved much the same result by simply selling off some of my tech holdings instead. But it's more interesting this way).

2 users thanked Tim D for this post.
King Lodos on 21/05/2018(UTC), Aminatidi on 21/05/2018(UTC)
King Lodos
Posted: 21 May 2018 00:53:04(UTC)
#22

Joined: 05/01/2016(UTC)
Posts: 2,967

Thanks: 686 times
Was thanked: 4578 time(s) in 1773 post(s)
Well I think getting towards the end of a market cycle, and with rates putting pressure on valuations, this year's been quite defined by a move to quality .. Lindsell Train especially has kept climbing – to now where they're about 3rd top performing managers in the UK out of 1,600. (and interestingly you're not really seeing that reflected in the ETF)

With rates climbing, you'd still expect the swing back to cyclicals at some point, but my bets on LT and LT-type stocks have proven one of my better decisions so far .. And Tech's still doing great too.

I still like quality – I think my thinking's been that this year would be volatile; would be difficult to time; and quality at least means you're not left holding junk if you find yourself with a big dent in the portfolio .. I'd just go the Lindsell Train route rather than the ETF, because I think the academic side of investing can be a very blunt knife
3 users thanked King Lodos for this post.
Tim D on 21/05/2018(UTC), Aminatidi on 21/05/2018(UTC), North Star on 22/05/2018(UTC)
Tim D
Posted: 21 May 2018 15:22:32(UTC)
#23

Joined: 07/06/2017(UTC)
Posts: 447

Thanks: 1638 times
Was thanked: 669 time(s) in 294 post(s)
This got me doing some number crunching on the full holdings CSV files for iShares quality and vanilla MSCI World (tickers IWQU and SWDA respectively for iShares ETFs; the IFWQ mentioned previously is the GBP denominated units of IWQU).

Interesting factoids:

* Discarding cash chaff, IWQU has 305 holdings to SWDA's 1655. The IWQU tickers comprise 25% of SWDA's market cap. Or put another way, construction of IWQU starts by throwing away 75% of the stuff in SWDA (which seems good... given the original appeal was the idea of "only buying the good bits of the index").

* Comparing IWQU weight with SWDA weight allows the degree to which the stock is overweighted vs the vanilla index to be determined. Top by overweight factor are:

Quote:
ADMIRAL GROUP PLC : ADM (United Kingdom, GBP) : 15.0
PARTNERS GROUP HOLDING AG : PGHN (Switzerland, CHF) : 13.0
SINGAPORE EXCHANGE LTD : S68 (Singapore, SGD) : 13.0
CI FINANCIAL CORP : CIX (Canada, CAD) : 12.0
HONG KONG AND CHINA GAS LTD : 3 (Hong Kong, HKD) : 11.5
SCHRODERS PLC : SDR (United Kingdom, GBP) : 11.0
SEI INVESTMENTS : SEIC (United States, USD) : 10.5
HARGREAVES LANSDOWN PLC : HL. (United Kingdom, GBP) : 10.5
SSE PLC : SSE (United Kingdom, GBP) : 10.4
RED ELECTRICA SA : REE (Spain, EUR) : 10.0
CENTRICA PLC : CNA (United Kingdom, GBP) : 10.0
HONG KONG EXCHANGES AND CLEARING L : 388 (Hong Kong, HKD) : 9.4
3I GROUP PLC : III (United Kingdom, GBP) : 9.3
T ROWE PRICE GROUP INC : TROW (United States, USD) : 9.1
EAST WEST BANCORP INC : EWBC (United States, USD) : 9.0
PINNACLE WEST CORP : PNW (United States, USD) : 9.0
MARSH & MCLENNAN INC : MMC (United States, USD) : 9.0
HENDERSON LAND DEVELOPMENT LTD : 12 (Hong Kong, HKD) : 9.0
RENAISSANCERE HOLDING LTD : RNR (United States, USD) : 9.0
CLP HOLDINGS LTD : 2 (Hong Kong, HKD) : 9.0


* However looking at the factors for the top holdings in IWQU, the degree of overweighting vs the vanilla index is much less:

Quote:
MICROSOFT CORP : MSFT (United States, USD) : 2.3
APPLE INC : AAPL (United States, USD) : 1.6
JOHNSON & JOHNSON : JNJ (United States, USD) : 3.4
AIA GROUP LTD : 1299 (Hong Kong, HKD) : 7.4
EXXON MOBIL CORP : XOM (United States, USD) : 2.2
ROCHE HOLDING PAR AG : ROG (Switzerland, CHF) : 4.7
MASTERCARD INC CLASS A : MA (United States, USD) : 3.6
VISA INC CLASS A : V (United States, USD) : 2.5
ALTRIA GROUP INC : MO (United States, USD) : 5.4
NIKE INC CLASS B : NKE (United States, USD) : 6.0
3M : MMM (United States, USD) : 4.8
STARBUCKS CORP : SBUX (United States, USD) : 6.9
CHARLES SCHWAB CORP : SCHW (United States, USD) : 7.6
NVIDIA CORP : NVDA (United States, USD) : 3.6
NOVO NORDISK CLASS B : NOVO B (Denmark, DKK) : 5.7
BOOKING HOLDINGS INC : BKNG (United States, USD) : 5.1
BLACKROCK INC : BLK (United States, USD) : 7.5
CHEVRON CORP : CVX (United States, USD) : 1.8
UNILEVER DRC NV : UNA (Netherlands, EUR) : 4.9
TEXAS INSTRUMENT INC : TXN (United States, USD) : 4.0


Which seems to confirm the idea that the megacaps bludgeon their way into the quality index more by sheer size than actually being all that "top quality" (and if I really believed this metric has value... then I really ought to be betting purely on the high-scoring stocks in the former list rather than buying the bunch of big but relatively mediocre-scoring behemoths in the latter).

Just to visualise things, here's a scatter plot of IWQU's holdings by their weight in IWQU and the "vanilla" SWDA. Note semi-log scale. Some quantization in the low end due to limited precision of iShare's data. The more extreme tickers (top 10 in IWQU, SWDA or by weighting multiplier) are labelled. Think it does illustrate how the stuff getting most heavily overweighted is largely in the smaller capitalisations (which could get you into a discussion about what factor you're really getting exposure to).

Quality vs vanilla scatter

Agree it's probably a "blunt knife" (nice metaphor). Will be interested to see how it holds up in a downturn.
3 users thanked Tim D for this post.
King Lodos on 21/05/2018(UTC), Aminatidi on 21/05/2018(UTC), North Star on 22/05/2018(UTC)
King Lodos
Posted: 22 May 2018 03:54:56(UTC)
#24

Joined: 05/01/2016(UTC)
Posts: 2,967

Thanks: 686 times
Was thanked: 4578 time(s) in 1773 post(s)
That's some really good work .. I'd think if you've got great analytics skills, you should be looking more into finding your own factors and inefficiencies.

a) I find the level of maths and analytics in investing surprisingly low .. Consider how long it took Fama and French to even demonstrate the Value factor .. In the academic space around investing, I feel like a giant .. In machine learning, I feel like a remedial student (so it's not just me).

b) Every quantitative strategy has a Best Before date, and it's usually before you've read about it somewhere .. 'Alpha' by definition is an edge that can't be attributed to something known, and I think that's telling .. What makes Lindsell Train interesting to me is that it's not a straight Quality tilt .. The process is clearly more driven by intangible aspects of a business .. The closest I've come to buying a Quality-style ETF is MOAT - VanEck Vectors Morningstar Wide Moat ETF, because that's doing something a bit more interesting .. But with a >0.5% fee, then the question might become: why not active??
1 user thanked King Lodos for this post.
Tim D on 22/05/2018(UTC)
Tim D
Posted: 22 May 2018 10:42:11(UTC)
#25

Joined: 07/06/2017(UTC)
Posts: 447

Thanks: 1638 times
Was thanked: 669 time(s) in 294 post(s)
Thanks for the compliment. In my previous 9-5 life, I did have a reputation for digging into things, taking an quantitative, objective evidence-based view and generally calling out other people's hype bubbles and magical thinking. Have also had occasional interest from city recruiters looking for numerate techies, based on my track record on various other forums. I can see the attraction. The amount of mathematics bought to bear across the whole field of investing does seem to vary enormously... I know someone who's a bond quant and they certainly employ some fabulously complex statistical modelling in their work. The real issue with diving in more myself is access to data (especially past timeseries for backtesting); haven't really looked into what's available but I assume it comes with a steep subscription price. One day I might get around to having a play on one of those crowdsourced trading algorithms sites (although I remember you pointing out some issues with them on another thread).

Yes MOAT has caught my eye too (I don't have any direct US investments - apart from one previous employer's shares - but I do visit etf.com weekly to check out all the weird and wonderful stuff they have over there, and Swedroe's articles and things on smart beta are always interesting). The problem with quantitative approaches is... you have to be able to quantify things. I suppose the hope with IWQU is that it's the existence of a moat is the sort of thing what sustains the above-average of Return on Equity and Earnings Variability metrics. But those are trailing information and it might take years for the disappearance of a company's moat to feed through to poorer numbers, whereas you'd hope an active manager (or Morningstar's researchers, in MOATs case) would see it coming before it happens. Somewhere I've seen some very interesting work on using things like metrics on patent filings to successfully predict future winners and losers (IIRC, sheer quantity of filing was a lousy predictor... think it was more based on keyword analysis to spot companies likely to move in on other companies turf) ... but like you say, by the time you've read about it it's probably too late.
1 user thanked Tim D for this post.
Aminatidi on 22/05/2018(UTC)
Aminatidi
Posted: 22 May 2018 15:21:00(UTC)
#26

Joined: 29/01/2018(UTC)
Posts: 220

Thanks: 389 times
Was thanked: 97 time(s) in 57 post(s)
Morningstar have "qualitative moat" on their stock factsheets.

I can't work out if I'm supposed to be paying for them or if they just don't have a clue what they're doing as if I simply Google "companyname fair value" it seems the first match is often a Morningstar PDF that I'm pretty sure is meant to be part of their paid offering?
King Lodos
Posted: 22 May 2018 17:29:20(UTC)
#27

Joined: 05/01/2016(UTC)
Posts: 2,967

Thanks: 686 times
Was thanked: 4578 time(s) in 1773 post(s)
Tim D;62661 wrote:
Thanks for the complement. In my previous 9-5 life, I did have a reputation for digging into things, taking an quantitative, objective evidence-based view and generally calling out other people's hype bubbles and magical thinking. Have also had occasional interest from city recruiters looking for numerate techies, based on my track record on various other forums. I can see the attraction. The amount of mathematics bought to bear across the whole field of investing does seem to vary enormously... I know someone who's a bond quant and they certainly employ some fabulously complex statistical modelling in their work. The real issue with diving in more myself is access to data (especially past timeseries for backtesting); haven't really looked into what's available but I assume it comes with a steep subscription price. One day I might get around to having a play on one of those crowdsourced trading algorithms sites (although I remember you pointing out some issues with them on another thread).

Yes MOAT has caught my eye too (I don't have any direct US investments - apart from one previous employer's shares - but I do visit etf.com weekly to check out all the weird and wonderful stuff they have over there, and Swedroe's articles and things on smart beta are always interesting). The problem with quantitative approaches is... you have to be able to quantify things. I suppose the hope with IWQU is that it's the existence of a moat is the sort of thing what sustains the above-average of Return on Equity and Earnings Variability metrics. But those are trailing information and it might take years for the disappearance of a company's moat to feed through to poorer numbers, whereas you'd hope an active manager (or Morningstar's researchers, in MOATs case) would see it coming before it happens. Somewhere I've seen some very interesting work on using things like metrics on patent filings to successfully predict future winners and losers (IIRC, sheer quantity of filing was a lousy predictor... think it was more based on keyword analysis to spot companies likely to move in on other companies turf) ... but like you say, by the time you've read about it it's probably too late.



Getting really good past data can cost a lot .. Occasionally I've resorted to things like bitmap to data convertors – if you can get charts (like on Gurufocus or Ycharts) with long-term metrics, it's perfectly possible to extract the data from images.

Sites like Portfolio123 let you build models with access to lots of data – and again, if someone were really technical, they could use Python and Chromedriver to build an app that spends all day web-scraping data off these sites .. Why don't I do it? I kind of agree the quant side is becoming too fashionable – therefore my instincts have been to go more towards business-focused investing (hence warming to Lindsell Train).

You only need to find 6 good businesses, after all .. There are no reliable trading strategies I can find that backtest better than that .. There seems to be a limit to long-term hypothetical returns of around 20%, which is about what the best traders and investors tend to do, without taking excessive risk.


The thing is I don't think fund managers are any better at understanding the brand value of Nike, Apple, Instagram, etc. than your average teenager (in fact, I imagine they're usually at a big disadvantage) .. So I quite like embracing investing as a way of looking at culture .. The quant side I think should be more of a checklist than a way of selecting investments.

I'm doing a lot with machine learning at the moment .. Stock market prediction is something no one's really cracked yet – not that it's necessarily possible/useful, but when you train a neural network on 50 years of S&P500, it tends to just average everything eventually .. If you imagine over 1000s of cycles, what happens between typical days, months, years, it winds up just learning the line the market's followed – rather useless .. I've mentioned after years of building algorithms to predict fund prices quantitatively (which worked out quite well), I realised the neural network in your head does a far better job, and I honestly don't think there's a better way to approach technical analysis than looking at charts, having looked at a lot of them
1 user thanked King Lodos for this post.
Tim D on 23/05/2018(UTC)
Dan L
Posted: 22 May 2018 18:23:33(UTC)
#28

Joined: 29/04/2018(UTC)
Posts: 6

Thanks: 2 times
Was thanked: 14 time(s) in 5 post(s)
King Lodos;62685 wrote:
I'm doing a lot with machine learning at the moment .. Stock market prediction is something no one's really cracked yet –


Just as an aside where do you get the cpu power for your machine learning? I am quite heavily into MS Azure and I can see the ability to open up things like machine learning and cognitive as a consumable service very powerful. One of the reasons I am invested in Microsoft
colin overton
Posted: 22 May 2018 18:43:16(UTC)
#30

Joined: 26/04/2008(UTC)
Posts: 258

Thanks: 51 times
Was thanked: 259 time(s) in 127 post(s)
What do you take "Quality" companies to mean?

A company that made a profit in the last quarter?

I used to work for a large American Corp. After a visit from "The Bains" an R&D colleague rang me up to ask what "a strategic business" was, as this is where the company was concentrating. I gave the above answer.
King Lodos
Posted: 22 May 2018 18:50:22(UTC)
#29

Joined: 05/01/2016(UTC)
Posts: 2,967

Thanks: 686 times
Was thanked: 4578 time(s) in 1773 post(s)
Dan L;62692 wrote:
King Lodos;62685 wrote:
I'm doing a lot with machine learning at the moment .. Stock market prediction is something no one's really cracked yet –


Just as an aside where do you get the cpu power for your machine learning? I am quite heavily into MS Azure and I can see the ability to open up things like machine learning and cognitive as a consumable service very powerful. One of the reasons I am invested in Microsoft


I'm just on a Macbook Pro, using Python, Tensorflow and Keras.

For image processing and learning large blocks of text, I'd certainly need a GPU or something cloud-based .. But I can fit networks like this for number sequence learning in under a minute – sometimes only a second:

(retina screen - apologies for size)
https://i.imgur.com/YVSS5pa.png

And in a way, I think with these kinds of problems, finding lower bandwidth solutions isn't always a bad thing .. I'm not sure there's any value in stock market prediction tbh .. I think there might be in feeding in lots of broad macro data (maybe month by month) and seeing if you can predict recessions – that kind of thing .. Lower bandwidth problems – rather than having a network look at what happened every day, for the past 50 years, 100,000 times, when 99.9% of the data will be noise
Tim D
Posted: 23 May 2018 23:44:17(UTC)
#31

Joined: 07/06/2017(UTC)
Posts: 447

Thanks: 1638 times
Was thanked: 669 time(s) in 294 post(s)
Fun stuff...

Thing is, you're up against folks like (as an extreme example) Renaissance Technologies and presumably lots of other quant hedge funds with access to vast resources and clever folks. That some random smart individual is going to stumble on the holy grail of "stock market prediction" seems unlikely (although the idea might form the basis for some kind of movie plot). However, I could just about believe you might pick up on something they've already discovered and are exploiting and enjoy it for a while too before it becomes more well known and/or arbitraged away (no idea how long that takes though... years, weeks, days, hours?). Good luck!

I was always very dubious about approaches which just look purely at the index numbers to predict the index. If that worked the AI algos should easily (re)discover, learn and successfully apply technical analysis ideas... but as you say they only rediscover the "stocks go up" principle. If there's an edge to be had anywhere, it surely ought to be from using additional data that noone else is... I'm thinking of things like the patent text analysis stuff mentioned previously, people using satellite imagery of walmart car parks to predict walmart earnings more accurately, investing in superior weather modelling to predict agricultural yields and commodity prices more accurately than anyone else, things like the "Market Prophit Social Media Sentiment Index" (or even that Phoenix trust's go-and-see approach to research)... stuff like that. The trend towards smart cities and open data might be a vein to explore, although from what I've seen of it apart from traffic data there's not that much available live or even with timely updates. (Glasgow was pioneering this a few years ago but I just looked at their data portal site again and I'm sure there's much less there now than there used to be).
King Lodos
Posted: 24 May 2018 02:49:05(UTC)
#32

Joined: 05/01/2016(UTC)
Posts: 2,967

Thanks: 686 times
Was thanked: 4578 time(s) in 1773 post(s)
This might make you optimistic ..

This is a VERY basic strategy I developed on Portfolio Visualizer years ago – I've posted screenshots from it before .. Incurs very few trading costs, and only uses ETFs.

It wasn't invented by messing around with numbers, but rather by having a theory on how money moves around – and while I can't give out the rules, I can say it's in Gold and Consumer Discretionary this month (so it hedges, it's not designed to maximise returns):

https://i.imgur.com/ZRKKpcH.jpg


What typically happens with quant strategies is they work amazingly in backtests, then horribly as soon as you start trading them .. Because backfitting is extraordinarily easy as rules get complex .. Often all you're doing is finding mathematical ways to describe past random movements.

But this strategy's outperformed more while I've been trading it (blue bars) – and I've overridden it on 80% of trades and done better (because I know why it works), but it does sometimes spot things .. and it's worked since the 80s – completely avoiding bear markets.


I studied Renaissance Technologies for a solid year .. In fact, they're not competition .. What they do is generate low but very stable returns – much like Blue Crest. (net 0% market exposure – everything's a pair trade, basically)

They generate 4-5% (or so) annually of raw alpha, over every asset in the market .. And what that means is they can use huge amounts of leverage .. That's where the return comes from.

There is the risk the fund could lose 18x more than it's worth – which could cause real problems, like Long Term Capital Management .. What their work is actually doing is taking that risk event and making it as SMALL as possible .. It's a mathematician's solution .. I think the chance of the fund blowing up is probably a less than 1 in 20,000 year event, and that's just smarter than doing what I'm doing, where you can underperform for a year .. They don't want to underperform for a month .. Understanding them helped me confirm my edge
1 user thanked King Lodos for this post.
Tim D on 24/05/2018(UTC)
2 PagesPrevious page12
+ Reply to discussion

Markets

Other markets