SVL-2000 Discussion of Results

Format for Printing

Format for printing

Request Reprints


By Gordon66
August 17, 2001

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

One year ago I initiated a public, real-time experiment to see whether it was possible to identify bubble stocks contemporaneously. In this post I suggested the experiment. In this post I finalized the components of SVL-2000 and SVL-2000m. And in this post I reported the final results.

For those into this sort of thing, it might not be a bad idea to go back and read the whole original thread. It makes for quite interesting reading given the events of the last year.

My motivation last August was partly intellectual curiosity and partly to get good tips on stocks to short. The experiment was triggered by my observation that a list of stocks with the "stupidest" valuations identified by us in November of 1999 (let's call it SVL-1999) did indeed go on to extremely poor absolute and relative performance over the ensuing 10 months. I wondered if that apparent success at bubble calling could be repeated.

Towards that end I orchestrated the construction of a set of hand-picked stocks to comprise a new stupidest valuation list, which I christened SVL-2000. I added to this a second set of mechanically selected stocks and christened that SVL-2000m. The criteria to create the mechanical list were abstracted from the criteria I and others had used to identify candidate bubble stocks (extremely high multiples, recent many-fold increase in stock price, etc.). Two stocks showed up in both lists (ICGE and CMRC).

I then set out the criteria for us to provisionally conclude that the valuations were "stupid" one year hence, which was for the SVL-2000 to seriously underperform the two most speculative indexes I knew of, QQQ and $GIN.X. I also said I expected the SVL portfolios would lose most of their value, but that they might not if the general market were to go up.

Now I will offer what I believe to be the four provisional conclusions we can draw from this experiment. I use the word provisional because any of these conclusions could be over turned if these stocks were to collectively beat the market by a substantial margin going forward (yeah, right�).

Conclusion #1: The valuations do appear to have been very stupid in an absolute sense.

The SVL-2000 lost 80% of its value, which means it would have to quintuple from here merely to get back to break even. Of the 12 stocks in SVL-2000, six lost about 90% or more of their value, and all but two lost more than 75%. I don't think anyone now expects the likes of LNUX, ICGE, CMRC, PALM, AETH, or EMRG to have any chance at a meaningful recovery, though anything is possible. Seven of the SVL-2000 stocks now have prices in the single digits, and four have prices under $5. I've heard it said that 75% of companies whose shares sell for under $5 are out of business within 5 years.

Six members of SVL-2000m (ICGE, CMRC, ARBA, INSP, JDSU, TIBX) lost 90% of more of their value, and the average loss was 84%. All of the stocks in this port lost at least half of their value.

Though there is a chance that one or more of the SVL companies could have a recovery miraculous enough to go on to long term market beating performance from a start date of August 2000, I think there is no chance whatsoever these stocks taken as a group will ever be worth even half what they were in August 2000 in risk-adjusted 2000 dollars. I think that is a pretty good working definition of "stupid".

Conclusion #2: The methodology I used to select the mechanical port is worth attending to.

The mechanical port out performed the hand-selected port. But more impressively, the four stocks in the mechanical port that lost the most were identified as the four stocks with the stupidest valuations by the mechanical process. Of the four stocks that lost the least, three were among the four stocks identified as least stupid by the mechanical process. The chance of this occurring by luck is very small. So let us revisit the logic behind the criteria used to identify the stupidest stocks in SVL-2000m:

(1) Higher P/S ratio (the higher the P/S the greater the effect of the inevitable P/S collapse)
(2) Higher market cap (the higher the cap the more difficult it will be to "grow into" the 100+ P/S ratio)
(3) Greater rise over the 52 week low (an indicator a stock that had the biggest run up)
(4) Greater subsequent fall from the 52 week high (an indicator of the stock that is exhibiting the down side of a bubble)

I think it may be worth dusting this mechanical screen off the next time there is a major general stock market bubble. See you in 15 or 20 years ;-)

Conclusion #3: The valuations were not just absolutely stupid, but also relatively stupid.

This conclusion requires more discussion than the first two. Yes, both ports did lose much more than every major market index including the very speculative QQQ. But it could be argued that the SVL stocks were more volatile than the typical stocks in any major index, even the QQQ, so in the event of a down market they should be expected fall further. Conversely, had the market had gone up, they would have been expected to go up much more, in which case today they'd look like a Smartest Valuation List (so this argument goes). Since the market didn't go up, we can't know that isn't what would have happened.

So to conclude they were relatively stupid--which actually is a higher standard in a severe bear market for speculative stocks than absolute stupidity--we have to look at the index with stocks that were most comparable to the SVL stocks, which is the $GIN.X internet index. This index fell 76% over the last year. SVL-2000m did indeed fall significantly lower than $GIN.X; it would have to go up 44% to be "only" as bad as $GIN.X. So SVL-2000m clearly qualifies as relatively stupid. Alas, the hand-picked list, SVL-2000, only a few percent more than $GIN.X, so that result at first glance seems inconclusive on the relative stupidity question.

But one thing to bear in mind is this: four of the SVL-2000 stocks were not internet stocks (PALM, LNUX, MSO, QGENF). If you were to remove these stocks the remaining stocks would have fallen an amazing 90% on average, which is much worse than $GIN.X. $GIN.X would have to lose another 58% to be off 90%. So it can be argued that the SVL-2000 did contain particularly stupidly valued internet stocks. (Think about that phrase for a moment. A PARTICULARLY stupidly valued internet stock�) Of the four non-internet companies, PALM (down 89%) has had the worst relative performance I can think of among the established hardware vendors, so it also seems to qualify as relatively stupid. LNUX (down 95%) has had much worse relative performance than RHAT (-82%) or CALD (-90%), the only other public Linux companies I can think of off hand, so that argues for it being relatively stupid as well. And QGENF (down 60%) did much worse than the closest thing I could find to a genome index, which is the Merrill Lynch Biotech Holdr (BBH) (-32%). That leaves MSO as being the only question mark on relative stupidity. Perhaps not coincidentally, MSO was the stock I identified as probably least overvalued at the time I created SVL-2000.

And if we look at the SVL-2000m list and take out the four health industry stocks (CRA, TIBX, MEDX, and HGSI) the remaining stocks, all internet related, lost 89% on average. So the internet stocks on SVL-2000m are also truly noteworthy bubbles in the vast sea of internet bubbledom.

Overall Conclusion #4: It is possible to identify bubble stocks contemporaneously.

Back when SVL-2000 was christened two complaints were made about the adequacy of the original SVL-1999 as evidence of the ability to call a bubble contemporaneously. Howardroark argued that the terrible relative performance of the 1999 might not be a credible data point on this issue due to the "survivor bias". That is, there may have been many such lists floating about here at TMF, and the only reason anyone came back to talk about this particular list is that it had such noteworthy performance.

It was also argued (by datasnooper) that the SVL 1999 was not 12 separate data points but actually only one data point because they were all internet companies. So, this argument went, we had not made 12 great calls, but at most one great call ("the internet is a bubble"). I said that was not exactly right, because several of the stocks on the list were not internet stocks, and also the port as a whole had lost much more than the $GIN.X internet index. But anyway, there was at least some merit to this complaint.

So the SVL-2000 experiment was designed to be robust to these two complaints. By making the call in advance and promising to come back in one year to judge the experiment, the first complaint is avoided. By including a mix of stocks other than just internet stocks, and by requiring that the ports underperform a basket of comparably speculative stocks, the second complaint is avoided.

Therefore, the poor absolute and relative performance of SVL-2000 and SVL-2000m are consistent with the conclusion that is it possible to identify bubble stocks contemporaneously. Is this conclusive proof on this issue? No, not in a academic or statistical sense. But as I said a year ago, there are certain things that intelligent, well-informed people can know are true that are not amenable to rigorous academic tests. And one of those things are, when the market has produced scores of small, unprofitable, unproven companies valued at several billion dollars and up, which is an accurate description of most members of the SVLas of one year ago, those valuations, taken as a group, are just plain stupid.


Ok, so are we all done? Has most the air been let out of the various bubble stocks in the market, or can we come up with a SVL-2001 list? Obviously there are far fewer stupid valuations out there now than there were one year ago. If you go to Excite to get companies with $1billion market caps and a P/S of at least 100 you get only 5 companies today, all biotech or genome companies in R&D mode, versus the 80 you got last year. On the other hand, as recently as three weeks ago KKD, a company that its own management has said will grow at about 25% a year going forward had a P/E of about 130. Its current P/E of 94 still seems pretty stupid. The rule-breaker staff thinks AFFX has a stupid valuation, so maybe that's a good candidate. What do y'all think--should we go for a three-peat? If so, volunteer your picks! BTW, in the interest of science I think we should only consider stocks with share prices above $10. So sorry, no loading up on penny stocks headed for oblivion.