« Me on CNBC | Main | New Release of Mint »
Latest Stories
- Excel Wankers and Recession Averages
- Sorry, New York is Closed. Check Back Later.
- Catching Falling 2009 Earnings Estimate Knife
- Survivorship Bias in Global Markets
- Talking Positions on a Lazy-ish Retirement Portfolio
January 28, 2007
Fun with IPO Data: Low-Rent Logit & the Letter "O"
This weekend's Barron's was an embarrassment of IPO data riches -- or at least it gave me something to do while sitting on a plane on Sunday. Either way, I got to thinking about 2006's class of initial public offerings, of which there were 181.So, let's wade in ...
Overall, it was a great year for taking risks on newly-issued stocks, with an average post-IPO return (the gains measured from an IPO opening trading until the end of the year) of about 25%, which handily beat the major markets over the same period. As has been well-documented now, technology did well, especially post-Labor Day, turning into the best IPO group of the year.
That's nice, but let's get to the meat of the matter. Since you and I can't buy every IPO, are there any metrics by which we might choose among them? And being fundamentally lazy people, we don't want any method that requires reading prospectuses, calling investor relations people, or reading analyst reports. Ick.
That doesn't leave us with much. To be honest, it leaves us with darn little, other than the names of the company and the dates on which they begin trading. Sure, you could fall back on things like sunspots and the like, but we're all market scientists here, damn it. So let's try a few things.
1) First letter of company name
Was there anything in a company's first letter that had a predictable relationship in 2006 with post-offering performance? You bet. According to my analysis -- okay, a quick-and-dirty Excel pivot -- the first letter "O", "F", and "K", in that order, were the top first letters for 2006 IPOs. Those three had a combined average return of 46.6% on the year, while the the rest of the alphabet -- filtering out of course, unpopular letters like "Q" that had fewer than 5 IPOs on the year -- had a combined average return of 22%
2) Company name length
Sometimes you just feel like you have to do a correlation. And the first one that came to mind was company name length and post-IPO performance. Did companies with more letters in their names do better in 2006 than those with less? Was there a correlation?
Well, the initial results weren't auspicious. The correlation between letter count and post-IPO performance in 2006 was 0.096. Disappointing, as you might imagine. While I initially hypothesized that investors would go for long names, figuring those ones had to be up to something mysterious and profitable, that didn't seem to be the case.
Or did it? Because being the long-ago lapsed researcher that I am, I decided to slice the name-length IPO data in a more self-serving way. Rather than running a correlation, how about a low-rent logit (or was that probit?), kinda. In other words, I tested the relationship between stratified letter count (under 20 letters, over 20 letters) and performance. Would there be a relationship.
You bet! There were 58 IPOs in 2006 from companies whose names were 20 letters or longer, and they had a combined average return of 24.8%. The remaining length-challenged companies averaged a paltry 13.5%. (Insert your own "size matters" joke here __________.) Net-net, you are better off buying the IPOs of companies with 20 letters or more in their names. Whoa, queue the biotech -onomics offerings.
3) IPO date
While I wanted to check returns by day of week, I can't figure out how to make Excel Pivottables do that. Sorry. Instead i was forced to look at monthly returns, which was still amusing. The upshot: The best month for IPOs last year was August, sporting a smashing 60% average return. The worst month, in case you're a minimax regret sort, was March, which had a 4.1% average return.
Sphere It
| 
Digg it
| 
Bookmark it
| 
Stumble it
| 
Facebook it
FYI, logit is used when the dependent variable (outcome) is binary, not the independent variable.
Still, thanks for the hard-hitting analysis. Showing again why the quants make the big bucks ;-).
I'm tempted to support the astrological argument (August 7, since you ask) but your data, while from a fairly small sample size, actually fits with what most professional money managers already know.
It all falls into that "buy on the AEA, sell on the H&Q" cliche you posted on a few weeks ago. Tech stocks have historically tended to do REALLY well from November to March. So if you are an investment banker trying to pitch an IPO, you usually can get the best price for your client (who is the issuer...NOT the buyer of the IPO) by allowing that seasonality to push up comparable valuations through the winter. Then when everything is at its peak in February/March you jam as much stock out there as you can. All of which makes March IPOs great for companies and bankers, but less good for investors...at least in the first year.
Contrariwise, the tech market usually bottoms in August. Valuations are down, fund managers are all on holiday, and demand is low. Which means that the company might not get as good a price...but the buyers of that IPO tend to do disproportionately well.
(By the way, buying in August for all tech stocks doesn't tend to be a bad idea. But it is safer with IPOs who usually only go public when they have visibility into the next quarter or two of revenues and earnings, and usually manage to avoid any nasty surprises for a little while. This is in stark contrast to public companies, who notoriously miss Q3 (September 30) earnings due to the summer slowdown and weak European sales.)
Duncan
I'll pile on.
Maybe you should read your own posts more carefully:
You could get the day of week by adding another column =WEEKDAY().
There is a way to do this directly in the pivot table although I normally just add another column to the data itself.
Having already concluded that IPOs which are Leos are the best, keen observers want to know if the astrological signs of the CEOs or founders also matter.
C'mon Paul, it's in the stars!
Ooh, and how about ROR by geographic location?
Putting off work,