Eddie, if you compute a p/e of a cyclical company using peak earnings, you will get a low value that is a result primarily of computing e at the peak of the cycle rather than averaging over a cycle. And that is what you are doing with CMOS and other stocks in the sector, regardless of whether you use last 12 mos, estimates for the next 12 mos., or whatever. It's just such a short period that you aren't averaging over a meaningfully large portion of a business cycle.
If instead you were to average the earnings over, let's say, a four-year period, and put that into the denominator of p/e, you would get a p/e that is more meaningful, showing the value that the company can accomplish over a cycle.
In the case of CMOS, operating earnings over the last 12 mos. is 5X larger than annualized operating earnings averaged over the last 4 years, as you can compute from the data in the table I made, below.* Therefore your p/e will be 5X higher if you average over the cycle, as I am suggesting. Instead of a p/e of 4.5, you end up with 22. That doesn't look ridiculously cheap anymore, does it?
CMOS YEAR ENDED OCTOBER 31,
--------------------
(in thousands, except per share amounts) 2000 1999 1998 1997
--------- --------------- --------------
Operating income (loss)................... 209,820 (4,013) (42,572) 15,063
4-year average ................................ 44,500
ratio of peak earnings to cycle-averaged earnings: 210/44 = 4.8
* I'm using operating earnings, rather than net earnings,
because that's more nearly what you were using, and it
isn't skewed by the interest income. |
