How Green is our Valley ( ie CAPS) : A journey in Sentiment Analysis
I started this last weekend – couldn’t complete – also I was expecting an impending reversal or “corrective” – so we can maybe add some additional datapoints. And there were some serious data challenges with the MF Screener
This is a bit of an off-shoot from my previous post
“Is a CAPS based Fund plausible? Attempt at Quant based strategy”
First of all – do not get too much worried – this is January – as usual new year enthusiasm and all – I’ll not be furiously blogging. Reasons as follows:
(b) Lack of time
(c) Lack of interesting topic
Additionally, I will be gone for a while in Februrary and won’t possibly be accessing CAPS.
Anyway the previous post was essentially a “Long Oriented” or Bullish leaning study – to understand what could work from a CAPs rating perspective. As I was doing it – especially in the backtest of the 9/19/2008 Long portfolio – I noticed that I could only find 27 stocks while I found about 110 in early March ie Fools rated more stocks positively in late Feb – not so much in Sep 2008. This got me thinking – could we use CAPS rating/bias towards understanding market direction?
Binve , the smart guy he is – posed the counter question – can there be a Signal – Bearish ( You know his slant : ) ) – I will focus on the signal part.
So here it is – my study ( You draw the conclusions – and it should be pretty obvious its no Holy Grail – well if it was – I wouldn’t be blogging about it – would I?)
(a) CAPS Bullish/Bearish Index
(b) Compare Contrast with general Investor Sentiment Data
A. The CAPS Bulls/Bear Stock Rating Index
Looking at the table –you’ll see its pretty straight forward –
1. Take all the Rated Stocks Day 1 of the month –
2. Classify 4+ as Buys < 3 as sells and 3 as holds.
3. # of Buys/# Sells is the Bull/Bear Spread
Pretty easy: Does CAPS or Can CAPS react in anticipation of moves – and on a comparative basis to the general investing public ( mentioned here very affectionately ..ahem as the “herd” or “sheeple” etc)
Issues with the Data: I had to hand run everything – making it tedious – the simple fundamental problem with this kind of data is “survivorship bias” – I was hoping to avoid dealing with it – however CAPS forced me to . In the month of April 09 – about 80+ stocks disappeared as per the CAPS database in terms of rating – it’s an error – the stocks in this list are like TOL etc…..and then in July the # of rated Stock count goes up by about 90 stocks! The set is not 1-to-1 swap – but most stocks comeback in the interim period – however majority was in that month. ( 4C Controls one of our old Red thumb favs amongst them) . I really had to take a hard look back to figure out how to adjust – the period is very critical – the first rally and the first corrective happened in these months. Anyway I simply forced survival in these months – ie ensured that the denominator of the index was consistent.
I wish I could say “TADA!!”. Well it opens up more questions and thoughts of introspection . My basic observations are on the chart – so I will not repeat most of it.
Generally CAPS Stocks ( read the distinction – since this is stocks rating based – its not CAPS members – the stock ratio which is being measured) – just went for a toss mid-2007 and looks like has remained so.
However – this Diffusion index as I see it – is essentially a Delta of 100 – ratio of Buys/Sells . So a negative number would have meant that there were more sells than buys – this has NEVER happened in CAPS. Even at the very bottom in April 2009 – Buy rated stocks held a small advantage.
(i) CAPS has maintained a fav or “select few” list of buys thru this period …this is the main reason why I believe the Buys/Sell ratio has not climbed all that much. Not many new entrants and old sells still languishing
(ii) Because if you care to click the Skeptic charm in CAPS – it shows about 2140 survivors – of about 70,000 players ( and some of them could be skeptics red thumbing Ultra etfs) – that’s less than 5%. Additionally the 100th percentile CAPS Score threshold has steadily moved up – when I started – about 1800 CAPS points meant you were Lord Gallagher or something – that number has edged up to about 4600 I think now.
(iii) There’s no doubt CAPS promotes ( in the sense of rating etc) – ultra speculativeness. And then due to accuracy – it promotes ultra egotisticism and adamancy. I do not need to elaborate too much ( you do not need to look further – looks at one of my picks called PALM – I wish it provided much needed shade to me!!!)
(iv) Well you can say this makes no sense – yes and no – because CAPS has a defense – Stock ratings are weighed towards Star Player picks….but Star players can maintain fiefdom thru the Accuracy counter ( my dear friend Tastylunch is unabashedly doing this currently for survival – and so do I – because mine’s pretty low in any case). This has resulted in non-closure and maintenance of old Red Thumbs
(v) In a perverse way CAPS scoring system – encourages the ULTIMATE - ACT & HOLD strategy. If you manage to catch a stock at the bottom – and it s a multi-bagger and good chance of not faltering – do not close it ever – the incremental score contributions of those picks can be astronomical. We all have learnt this first-hand haven’t we since Mar 2009?
On the other hand if you manage to Red Thumb at the peak – and that stock fell 70/80% - you can easily stay put even in a bull cycle – because you will lose points – but unless the stock registers close to an all-time high – marginal effect on your accuracy is zilch . There are a lot of players like this – especially the dormant – “open and shut” inactive players.
SO WHAT”s USEFUL? ( Things to watch out for)
Its always the marginal movement – that’s what counts – not the absolute most of the time.
(a) Bearish Signal: Another silde in the index - Run for cover – especially new low
(b) Bullish Signal: If the index breaks above the Aug 2009 high of 4.1% - significantly – ie decent jump. CAPS is not so good at timing – that should be obvious – but I think it would be a nice – directional indicator
The above statement is extremely opinionated and judgmental – there is no DATA/BACKTEST to prove this. So take it with whatever skepticism you can muster.
THE EGO-BUILDING PART: Comparison to general investment sources
Coastaltrader (CT as we call him in CIL –coveritlive.com) – pointed me to this Hewitt ( I am sure a lot of your own Company 401Ks are administered thru them ) site – where they publish data around 401Ks they manage. It’s a good side check it out. (http://www.hewittassociates.com/Intl/NA/en-US/OurServices/IndexObservationList.aspx
What we wanted to understand was the extent of the so called “money-on-the-sidelines” [ There is definitely some element of truth to this – all the CAPS surviving bears – can’t or couldn’t have survived in the real-world without covering – but given their slant – my bet would be 100% cash for them. BTW – if you were wondering – I was a restrained long-especially post the Jul09 S&P 905 and currently just about smidge long – mostly cash]
Anyway , here’s what I did – another very painful round of data entry – I am afraid. Took the monthly data available from Aug 1997 to Dec 2009 and smoothed it in Excel and plotted it ( For definition of the index look here: http://www.hewittassociates.com/Intl/NA/en-US/OurServices/AtAGlance401k.aspx)
You have to understand this index – its truly the “traded” portion of it – ie the “movement” in 401Kfunds. Dormant money doesn’t play any role in it. So % Fixed Income days – is simply count of days as a percentage for the month – when more money ( trades) flew into Bonds
Its much better if you smooth the data for 3SMA and plot it against the S&P.
(i) There is some “Negative Divergence” ( TA term) going on – since 2003 between % trades in and out of the 401Ks. The bottoms steadily rose thru 2003-2007. So people who trade or rebalance 401Ks became steadily more defensive
(ii) The opposite happened thru the bear – less and less people traded into Bonds with each leg of the Bear!!!
(iii) Since Mar 2009 – there has been not a single “flight-to-safety” period ( going above 60%) – people have hung on – similar to the 2003 late period.First sign there was 2004. A scenario likely to repeat or possibly repeat in 2010 – IMHO!
(iv) However the KEY thing is that - its based on trades….Very few people do that. Here’s the plot of these “movements” as a % of Total balance on a Monthly Average
Very few people touch their 401Ks…..if that happened – Markets will move in a way – which is UNTHINKABLE!!!
Thus 401K movements mainly come from 2 sources
(a) Trades or Rebalancing
(b) New money/Contributions
So the second piece – is the contributions. There are 2 available – one is the all in ( including Employer match ). Since most employer match is equity – I focused on the employee piece of it. You may very well look at the total balances – its available also – I personally believe – its slightly auto-correlated….since most people hold – and if they follow a decent balanced allocation – Bull Cycles will increase Equity Balance portions and Bears deplete the same – I have a bunch of comparisons – keep that thought in mind.
Here’s the current last 2 months pic
We are currently seeing about 27-28% of the “new” money going to “Safer” instruments. Lets look at some trending of this
Peak 401k Bullish Activity month :Feb 2007
It was around 21% - and almost the same in Oct 2007 – market peak month.
Now compare to the “scare the bejesus “ months ie
So about a 14% ie 66% additional movement. At current level – we are about 50% - almost midway between peak to trough allocations. Pretty much the same retraces in Indices also – (Strange isn’t it?)
Now I tried to compare to the last bear ie the 2000 tech crash. The allocation data is not available there – but the balances are – and its good because it provides a view AT THE TOP!
Ok ! Ballpark same number 22% - same as the 2007 early allocations – which means 80/20 rule applies here – people aspire to that Balance.
Next is Feb 2003 ( bottom)
And we also have the comparable allocation ( ie new money) here
Ok about 38% in Balances here – and 28% Allocation. (Comparable to the 35% allocation in Mar 2009). Seems to me there’s a rhtym of 7% cycle here…..2009 was clearly 1 amplitude higher!!!
We can also compare the Balances for 2007 and 2009 – here they are
And the balances as of Dec 2009:
The total balance allocation to cash/bonds went to 44.5% ( about 7% higher : ) than 2000 bottom) in Mar 2009. And it’s back – obviously 50% from there – to about 34%
Last section – can any class actually ANTICIPATE?
First a case in White-Noise or randomness….. S&P Index and 401k Index delta’s plotted together.
Its clearly random…..Ran the correlation analysis at 1 interval lag ( ie to see predictiveness)….Here’s the results for 401K “Traders” and CAPS “Raters”
Both are negative – ie counter-trend anticipatory – the 401k is basically useless <5% significance. The CAPS one- its better than nothing – but wont’ stand a model productiveness test – about 14%
Let me start with one chart borrowed from GV ( GoodVibes – www.iamgv.com) blog
(a) It shows investor – I call “Trader” sentiment. Clearly its bullish
(b) The 401K Trader Index is also basically “bullish” –since no one essentially seemed to have abandoned stocks – during both the Jul and Oct corrections. Usually corrections are marked by at least 60% trading into bonds etc month!
(c) CAPS All-Stars are BEARISH – on a stock rating sense. Its calling this a Bear Market Rally
(d) Overall CAPS players I guess are Bullish - <5% Skeptics currently
(e) Money on the sidelines and “new money” contributions: This is neutral – 50% retracement towards peak – in line with Index retraces
(f) Also based on a Ultra-Bullish allocation of about 21% cash – there room to go as far as 401K investors
My feeling is there will be correction here ( possibly started) which will force that move to “safety” – but given the allocations – there seems to be a distinct semblance of truth about money in sidelines.
Should be pretty obvious this was a painstaking blog. It would interesting to see the number of recs :) – LOL!