March 17, 2008

First, Let's Kill All the Data

(c) copyright ,View from Silicon Valley, 2008.  All rights reserved.



We don't presume to write Shakespearean drama (or is it tragedy?) but one interpretation of the "Henry VI" line, "First, let's kill all the lawyers," was meant to say the best way to cheat people was to kill off all the lawyers first.  An alternate interpretation was to distinguish between the "law" and how "justice" is unfairly administered when lawyers get involved.  For the purposes of today's discussion, either interpretation will do!

We believe the best way to cheat people is to deprive them of objective data.  At risk of going a metaphor too far, "Something is rotten in Demark."

By now we all know about machinations with the CPI and the Fed's decision to stop publishing M3.  Not only is this "old news," but you can read about it in a lot of other places.  We prefer to focus on Silicon valley-centric issues.

In this piece we will highlight a slightly broader cross-section:
1) There is a wide variety of pieces by folks selling stock-picking services.  Their agenda should be obvious. 

2) Slightly less-blatant is saturation advertising by brokerage houses.  Their data doesn't need to be "right" so much as they need you to place your money with them while, preferably, frequently trading your positions.

3) Somewhat overlooked are free columns which slip in tidbits about how "hedge funds and other alternative style investments" are essential to solving the day's "crisis on our long term returns and retirement portfolios."

In addition to getting slanted information, there is the mis-use of terms, such as:

4) The local paper publishes "Top Traders" in the business section, listing a cross-section of publicly-reported sales by Silicon Valley company insiders.  Problems with this info include:
        4A) the so-called "traders" often do not risk a nickel of their own money;
        4B) sellers choosing to avoid public scrutiny can just buy a "collar."  Our favorite was a top guy at JDSU who openly wrote "calls" against most of his options.  He then used funds from the "calls" to buy "puts" on JDSU stock.  He then delivered stock from his options/grants against in-the-money "puts" as the stock fell.  This was in no way illegal.  It was just a lucky coincidence he wasn't required to report his insider stock sales since he was technically just delivering on the "put."  (In other words, we can't judge an absence of insider sales as any type of endorsement for that company.)

5) Probably the first instance of "missing" data was when Doug Noland 's Credit Bubble Bulletin began to change the format, or outright omit, year-over-year mortgage growth.  Even as housing softened in late-2006 and into 2007 y-o-y mortgage growth stubbornly stayed in double-digits.  About the time the y-o-y figures might be expected to finally(!) start dropping, the data disappeared for several weeks and we quit following it.  This was such a small bit of "missing" flotsam in a sea of data that we didn't even follow-up with Prudent Bear at the time.  We figured their source stopped reporting it and/or they had bigger fish to fry.

6) SMG (Silicon Manufacturing Group), a sub-group of Semiconductor Equipment and Materials International  (SEMI), stopped breaking down wafer types, including EPI wafers, beginning 1Q07.  (EPI refers to "Epitaxial" circuits, whose speed and power characteristics are best-suited to yield well on this type of wafer.  Conversely, "polished" wafers offer a higher-purity silicon suitable for a wide variety of popular chips.)

EPI wafers are used to build things like MOSFETs and other power management chips.  Now power management is a nice, and generally profitable, segment of the semiconductor business.  However, this segment tends not to grow at the same rate as Processors, NAND flash or DRAMs.

If EPI wafers lost or gained share among wafer inches shipped in 2007, it becomes  a multiplier (or divider) against the implied over-production of wafer inches vs. IC sales that we track quarterly. (see (http://www.viewfromsiliconvalley.com/id71.html).  

Again this seemed like a minor change and we glossed over it at the time.

Then the issue got a little more serious. 
7) Late in 2007, DataQuick (DQ) slowed their weekly on-line update of local housing data.  Instead of uploading most weeks by Wednesday or Thursday, the data started not coming until Saturday.  This wasn't too awful since we could just wait until Saturday to add each week's update to our on-line compilations,

This change fully achieved "minor annoyance" (to both us and our readers) but falls short of a major red flag.

8) Early in 2008, the local paper re-formatted their weekly DQ publication.  Instead of putting DQ's real estate sales statistics and map in their real estate section (duh!!), they moved the table to the "Home and Garden" section?  However, the data on number of listings, mortgage rates, percent of houses selling above list, etc. remained in the real estate section??  To save both sets of data, we now have to find and cut out pages from two different sections of the paper.  This progresses to a more significant annoyance.  Our paper has been fished out of the recycle bin a couple times already this year.

We can only imagine if this move is at the behest of the paper's real estate advertisers and/or DQ.  Now that the weekly DQ map shows broad areas of y-o-y price declines, keeping such info separate from new listings, recent sales and mortgage data is relevant.  After all, we can't have prospective buyers asking inconvenient questions about why their agent told them to rush out and buy since  Silicon Valley real estate prices never go down.

If you still wondering if this is really true, please check out 1989 to 1989 prices for San Jose  in, "Myth #2: Silicon Valley House Prices Never Go Down"  (http://www.viewfromsiliconvalley.com/id315.html).  Buyers in 1989 were underwater until 1998, or about the time of the Netscape IPO.  Anybody see another dot-com boom coming along to sustain /grow current Silicon Valley real estate prices?  (See upcoming missive on IPOs.)

9) DQ has now further reduced the frequency of their "weekly" updates on their web site.  As of this writing, DQ's web site shows data only through January 28 (48 days old), while the data extracted from the paper goes up to February 26, 2008 (18 days old).  If you're selling real estate in a declining market, "old" data is "good."

This advances to the level of active irritant.  Of course, if you read "View from Silicon Valley" every week you get the newer, and "better," data.  ;=)

10) Just to round out the picture, a couple weeks ago we noticed a change by the Semiconductor Industry Association (SIA).  We have long railed against SIA's proclivity to write headlines based on "three-month averages." SIA claims using "three-month averages" instead of "actuals" smoothes out month-to-month "noise" in SIA members' sales.   This is self-serving nonsense!

Serious investors track year-over-year changes.  Y-o-y "actuals" highlight trend changes while a "three-month average" is worse than useless for this purpose.   To state it more clearly (or redundantly), using "averages" disguises when SIA sales figures turn --up or down. 

When it's our own money at risk, we want to be first in identifying a change in trend.  Waiting two or three months for alleged "noise" to cancel out of "three-month averages" is not conducive to reaching that end.

10A) Not only did SIA not include "actuals" in their last press release, their web site now shows NO prior publications of "actuals"???  If there was any formal explanation from SIA, we haven't found it.

10B) As an aside, we have "actuals" back to 1976 and could extract each month's new "actual" from each new average.  Unfortunately, running a free web site means the bandwidth is limited --unless somebody wants to start paying?  (Hint, hint...)

If you follow SIA data, or invest in companies whose stock prices leverage off of semiconductor sales, switching from "actuals" to only "three-month averages" is a big deal.

Conclusion:
Are any of these changes a huge, catastrophic portent of doom?  ("Get thee to a nunnery?")  No, of course not.

However, these changes do imply the underlying markets (mortgage growth, wafer area, Silicon Valley real estate and SIA sales) are in turmoil.  If nothing was "wrong," why would the respective parties risk drawing attention to their numbers by changing their format and/or frequency?

One obvious answer may be that things are going badly wrong in several of these areas.  Killing (OK, hiding) the data is the quickest /easiest way to disguise the trouble.

Last pun: "Oh data, data! Wherefore art thou data?"