Forums  > Trading  > Value-Quant research/literature?  
     
Page 1 of 2Goto to page: [1], 2 Next
Display using:  

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-03 21:26
Hi NPers,

I was wondering if you could recommend research/literature relevant to a "value investing with quantitative insights" investment style. I suppose an example of such a style would be the ubiquitious "initial quantitative screen" which filters an initial universe of gazillion stocks to a more manageable number for in-depth analysis. But I'm sure the NPers have come across more interesting stuff?

(My personal interest is in the more arb-y stuff like statistical arb, but the long-biased shop i'm interning at has tasked me to look into creating a scary goldblum-esque "value-quant" model...)

I've been scrounging the campus library, and only came up with Frank Fabozzi's "Active Equity Portfolio Management"... Confused



Many thanks,
Rogaine


Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-04 15:41
On second thought, my request does seem rather broad and nebulous. None the less, I am interested to see what relevant stuff NPers come up with! Big Smile

Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

IAmEric
Phorgy Phynance
Banned
Total Posts: 2961
Joined: Oct 2004
 
Posted: 2006-02-04 15:51
Never hurts to start at the beginning.

ball_lightning


Total Posts: 295
Joined: Dec 2004
 
Posted: 2006-02-04 17:46
I have tried using an initial screening based on fitting a relationship between P/E versus ROE or P/BV versus ROE for a particular industry segment. This is preceded by a scatter plot to see if such a relationship exists at all and whether a linear or non-linear regression is required.

Once you get the relationship rank the companies in terms of the degree of deviation from the (industry) fitted line. Then if you are a long only shop go for the apparently undervalued companies for a detailed analysis.

Is this quantitative enough to start off?

uNclearPhynance: Illegitimi Non Carborundum

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-04 21:18
Eric,
right you are. In fact, I spy a well-thumbed copy of Fisher's "Uncommon Stocks..." on the bossman's credenza as I speak... Smiley


Lightning,
thanks for the ideas. I did however come across similar ones in O'Shaughnessy ("What works on wall street") and English ("Applied equity analysis"). I also note that when O'Shaughnessy tried to apply his ideas in a fund, the results weren't too hot... Confused

The key I suppose (and as you have noted) is that they are mere initial screens. I guess the more quantitative long-biased shops (Citadel?) would have screens similar in spirit, but with considerable more nuance (e.g. not just P/BV but P/(quality of BV)...). Nonetheless, after the screen has spat out its results, it would seem that good old security analysis rules... Blendertime


Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-05 09:27

For any quantitative screening process based on accounting data, the first step is to make sensible adjustments to the accounts. I'm talking about the usual stuff such as use of historical cost in BV, capitalising leases, treatment of merger accounting and so on.

 


The night is young, the mood is mellow There's music in my ears

kr
Founding Member
NP Raider
Total Posts: 3561
Joined: Apr 2004
 
Posted: 2006-02-05 12:20
Of course I am a big fan of 'starting at the beginning' as well, but you have to appreciate the cyclicality of fashions in all this.  The trend of what people want is always shifting, with the ratios wandering along with it (ok, this is back to the beginning as well, Keynes, wasn't it?) 

Whether there is some decipherable logic to the wandering is tough to say.  It is not enough to get the structure right but the timing wrong obviously.  Whether one can quant up the timing aspect is a good question that easily runs into the issue of stationarity. 

So one can hope, but there are limits.

my bank got pwnd

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-05 20:00
Thanks for the insights.

Following Johnny and KR's musings, I shall duly bone-up on my accounting Cry and Niederhoffering (law of everchanging cycles etc. etc.)...


Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

shavinOccam


Total Posts: 210
Joined: Oct 2005
 
Posted: 2006-02-06 09:32

"arbitrage price theory", and then there is a guy called robert haugen, who published something late nineties on multiregression on balance sheet data. i fell over him because one booklet of him was on the book shelf of bnp cooper neff, then one of the big companies doing quant stuff on equity.

AFAIK there is about a dozen of relevant non-price factors in equities. usual suspects. single most important seems to be earningsforecasts, best source for that is thompson (bought IBES) and i have access to JCF, who are cheaper. best balance sheet data on historic basis is compustat.

important thing to realise is that you have quite limited data over time, meaning at best 240 data points per stock per balance sheet data of interest, since monthly is best you can get (earnings forecast might be an exception). having clean data over 20 years will not happen on too many stocks, so you'd better be careful not to overfit by adding too many different factors - hence all kinds of ratios between basic factors fall into that camp as well. what you can do about is taking more priceDerived data into the equation. mas, vola, nothing necessarily fancy. plus looking at more stocks and moving to more transversal analysis over a horizon of maybe 1500 stocks, instead of looking longitudinal over time on a few hundred stocks. then you come to the key problem: survivorship bias. THE f...... most important problem to overcome.

short sales might be interesting as well, but i think it is hard to get historics and i think it is still just on nasdaq issues ...

BTW graham dodd ... just try to read this ... i promise you won't have sex for months ...


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-06 14:17
"Following Johnny and KR's musings, I shall duly bone-up on my accounting Cry and Niederhoffering (law of everchanging cycles etc. etc.)..."

yeah, it's unfortunate but unadjusted accounts numbers can be deeply misleading. But many many quant finance people never bother to learn this stuff and are surprised when they get bitten.

 


The night is young, the mood is mellow There's music in my ears

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-06 16:39
Thanks for the tips, Occam and Johnny.

1. Haugen: I'm reading through the Haugen paper ("Commonality in the determinants of expected stock returns" right?). Looks interesting. If any NPers want the PDF, please drop me a mail. Smiley

2. Data: I only have Bloomberg, which is great but also missing many datasets which I could use (e.g. Zinc prices in China, not just LME...). Wish I had Datastream/Compustat. I also wish I had a team of professional data masseurs to fumigate and common-size the data as per Occam/Johnny's suggestions. Cry

Okie... lots of food for thought for me to chew on. I shall endeavour to post in this thread any relevant papers/literature I find in the future.

Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

doctorwes


Total Posts: 577
Joined: May 2005
 
Posted: 2006-02-06 17:57
I have to agree with Johnny. It doesn't make sense to compare reported numbers from different companies when they are often based on different accounting conventions. In any good investment firm, the analysts spend a lot of time getting the numbers onto a consistent, comparable footing; Compustat is just the starting point. The whole exercise requires a strong background in accounting. If you don't do that, you can't even get started, in my view.



shavinOccam


Total Posts: 210
Joined: Oct 2005
 
Posted: 2006-02-07 09:51

thnx for these thoughts ... i had this problem with bloomberg historics ... i hoped so much to overcome it by using compustat, hoping that actually they would do that job ... wel, and celarly different jurisdictions are a pain. do you know sabre? they are trading some kind of balanceSheetDataStrategy and i did not see them as ... well ... too sophisticated ... though i might do them wrong ...

f...


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-07 10:25

ok, but this implies that I'm merely talking about some cleaning up job that could be easily automated. But that's not what I mean at all. I'm talking about proper intelligent adjustments for leases, for merger accounting, for non-balance sheet assets such as intellectual property. I'm talking about spotting P&L not taken through the P&L statement but buried in the notes. These are not things that you could expect Compustat to do. They are things that require the human touch. For once I'm in with AndyM's Promethean Spark campaign.

 


The sound of one bear, uh, in the woods

tripitaka


Total Posts: 982
Joined: Jun 2004
 
Posted: 2006-02-07 11:03

but you can't do all the analysis and adjustments for a whole universe of stocks, as I think that the initial post says. It can be a long and pretty painful process to do it properly and so you want to do it only where you have some confidence that it's worth the effort.

i guess it depends on your outlook. If you're a quant guy, you want to get the numbers right, so your data mining/model/whatever can go to work, and so adjustment comes first; if you're  grhame and dodd kinda guy, then you want some quant help to weed out the low hanging fruit (mixed metaphor) to minimise your wasted analysis.


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-07 11:24

It raises the same question as exists in many other parts of finance, which is how to go from cottage industry-style artisanal efforts to industrial production. I agree with you that it's not an easy question to answer. I don't agree with you when you say that this stage should be skipped.

A simple example. Consider three companies identical in all ways except that company A has no debt, company B has debt and company C has the same debt as company B, but predominantly in the form of leases. How should the share prices of these companies relate to each other? If they all have the "correct" share prices, which ones will appear in naive filters based on unadjusted accounts? How will you feel when your entire portfolio consists of companies with substantial debt in the form of leases when business slows down?

btw, in conversation the other day someone remarked "and in that episode Tripitaka became a woman". I bit my tongue, but it made me smile. What on Earth was that about?

 


The sound of one bear, uh, in the woods

tripitaka


Total Posts: 982
Joined: Jun 2004
 
Posted: 2006-02-07 11:40

agreed, but from my point of view, if all three are appropriately valued, i don't want to waste time working on them - as far as i can see, i can

a) analyse all three, plug them into the model, decide they're fairly valued. Time taken: 3, investment ideas: 0

b) plug them into the model, like the look of company C, work on that, adjust for the leases, decide it's not cheap. Time taken: 1, Investment ideas: 0

and B is the better result.


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-07 11:41

Fair enough. I guess that takes us full circle back to Rogaine's "ubiquitous initial quantitative screen".

 


The sound of one bear, uh, in the woods

shavinOccam


Total Posts: 210
Joined: Oct 2005
 
Posted: 2006-02-07 15:21

i'd say the typical pure quant approach would be: errors net each other out (i expect some rotten tomatoes here ...). by running several hundred positions you try to overcome that weakness. and then: it is the data pretty much everyone looks at, so you might have a mistake, but you're mistake is in line with the street.

sure, that is most stupid for single positions ...


Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-07 15:31

It doesn't make sense where there are systematic biases either. Like in accounts ...

 


The sound of one bear, uh, in the woods

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-07 18:39
Occam,

were you referring to the Sabre fund by Mathews Capital? Spectacular numbers they put up...



"It raises the same question as exists in many other parts of finance, which is how to go from cottage industry-style artisanal efforts to industrial production." Agreed. I guess by its very nature accounting data simply lacks a lot of qualitative content. Which is why analyst-monkeyboys are still in demand... Big Smile




Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

kr
Founding Member
NP Raider
Total Posts: 3561
Joined: Apr 2004
 
Posted: 2006-02-09 22:34

I think it's a mistake to even start thinking about industrial production.  You are never going to be perfect, so it's a question of the effort required to improve signal-to-noise.  Even saying it that way is pretty sloppy.  A way to approach it, which I think is fairly common, is:

- form a thesis about the economic environment
- establish which risk exposures you definitely want, and those that you definitely don't want
- figure out how that translates into quantitative, but mostly qualitative, boxes that should get ticked on a quick screen
- slowly comb through a list of possibilities, take the ones you like, and dump the ones you don't like
- don't be afraid to take small tangents
- BEWARE THAT THE ENVIRONMENT IS CHANGING AND THE THESIS MAY BECOME FALSE

Something that may not be so clear to people here is that a very huge part of this business derives from the new issue market.  In this context, a price is set once (at IPO), there are allocations that indicate where your demand curve is relative to the market, and secondary market considerations are minimal.  This is true on both buy and sell sides.  In this context, the thesis may be somewhat clear.  For instance, we are seeing a lot of jumbo leveraged loan deals, because PE firms have massive capital and feel that doing megamergers is good work.  It is a style, where the clients want it and the dealers are selling it, and there seems to be some [meager] consensus about the risk exposures to be concerned about on both sides.  Investors may plunge in headlong, find that the thesis switches - or realized events prove it wrong - and suddenly a style is dead. 

Being optimal on the numbers is not really how it's done.  On sellside, you are faced with a pipeline of deals, and get a chance to bet on completion.  At the end of the day, it's yes or no, and you try not to split hairs - because 'deals on the edge' are dangerous, possibly costing you a lot of time and then delivering you a failure.  Investors are bobbing along relative to their market competitors, and if they stray too far from the thesis, they will screw up their ramp... either they spend all their money at once, or they are all cash.  Ultimately this means your portfolio vol won't hit its target.

The thing I like about Graham/Dodd is that it's not really about accounting.  It's about simple companies, anecdotals, and commonsense.  I don't intend to be an expert on pension accounting, for instance, because it's dull and done in a way that makes my mathematical sensibility cringe.  If that's what investing is about, I quit because I have better and more interesting things to do.  These days, you are told that doing things in a simple way will get you killed by tricksters like Enron.  It could be true, but you have to ask yourself:  Would this kind of animal fit into your thesis?  Was their business plan, analyzed through a critical commonsense eye, attractive in any sense? 

Ok, at this point, 'industrial machinery' is not as computer-dependent as you were thinking.  No perpetual motion machines.  Industrial approach means a process for analyzing the qualitative stuff, in human (and not machine) time. 

BTW, Eric if you are still following this, I just received my copy of MJ Whitman's "Agressive Conservative Investor", which I think you might like as a sequel to G&D.


my bank got pwnd

rogaine


Total Posts: 67
Joined: Feb 2006
 
Posted: 2006-02-10 04:26
kr dude, interesting to hear your thoughts as always. As I understand it, you are advocating a macro/country/sector overlay/thesis + simple value-quant screen + heavy-duty qualitative (human) analysis. Sounds about right to me. More employment for dismal scientists! Applause

The Aggressive Conservative Investor is also on my reading list. Another book NPers might also consider reading is Investing with Young Guns by James Morton. Cringe-worthy title I know, but it contains interesting and detailed interviews with several value hotshots like Roger Guy and Wayne Cooperman.

Art Vandelay Capital Partners
* Plain Venetian Blind Arbitrage * Latex Swaps *

Johnny
Founding Member

Total Posts: 4333
Joined: May 2004
 
Posted: 2006-02-10 09:18

kr, I agree with much of what you have written, particularly your point about not hoping to be perfect. I was using "industrial" to refer to scale and not to quality precisely with the implication that diversification (i.e. scale) is the best answer to not being perfect.

To me it seems an interesting and useful question how to design an industrial scale investment process. Many small shops have artisinal processes dependent upon the skills of a small number of key people. Big shops tend simply to reproduce this artisinal model n times over and therefore fail to capture returns to scale. It seems that a trick is being missed.

 


The sound of one bear, uh, in the woods

misfit


Total Posts: 14
Joined: Feb 2006
 
Posted: 2006-02-10 12:23

you might find this useful(http://faculty.fuqua.duke.edu/~charvey/Teaching/Independent_Studies_2004/) and throughout the parent directory of that link for more.

btw, for some encouragement:

1. dont expect this to really work. it would be like winning at roullette using the rules of poker.

2. the databases your looking at have been probed by at least 50,000 other people.

3. dont put your own cash in this or any other "quant"/value fund who offers it.

 

Previous Thread :: Next Thread 
Page 1 of 2Goto to page: [1], 2 Next