When I was working for McGraw Hill's Electronics magazine in the early 80s we published a yearly market report that covered all significant product areas, which area was up, which was down, from the preceding year. It was not hard to track how individual market analysts took our data and presented it as their own, to which next year's Electronics report extrapolated the numbers based on the analyst' reports as well as our own research. Point? market analysis is part art, and part science; believe but verify.
Hi Peter - your comments are precisely why I run the Cowan LRA Model each month immediately following the release of the WSTS's monthly HBR (Historical Billings Report). This allows the model to capture the latest, up to date global semi sales (as well as any revisions from prior months)- the monthly forecast is therefore dynamic in the sense that it can "pick up" any of these "non-linear effects" that are sometimes superimposed on the "normally linear" sales activity of the semiconductor industry.
The obvious effect is that the Cowan LRA Model's monthly forecast numbers do not "sit still" but evolve each month as the latest sales numbers become available.
Regards, Mike Cowan
The latest Cowan LRA Model's forecast expectation for global semi sales came in at $299.8 billion. This results in a 2012 sales growth forecast estimate of 0.1 percent - essentially zero growth for 2012. These latest forecast numbers are based on the WSTS's May sales HBR posted on their website July 6th.
The Cowan LRA Model's forecast methodology employed in deriving these global sales and sales growth results exploits linear regression analysis of the last 28 years (1984 to 2011) of historical monthly (actual) sales numbers gathered and published by the WSTS. Consequently the resulting numbers are based on "facts" (actual monthly sales numbers) and are operated on by statistical analysis, namely linear regression analysis - thus the model's name LRA!
Mike Cowan (independent semi industry analyst and developer of the Cowan LRA forecasting model)
Note - for more info see the following URL = http://www.electronicsweekly.com/Articles/10/07/2012/54089/cowan-predicts-flat-2012.htm
Basically the economic trends of the semiconductor markets are complex and have a chaotic non-linear element to them, which makes their measurement in strict time units somewhat arbitrary.
This chaos is not least because markets are human-modulated entities and human beings are not rational and markets by their very nature have a herd-like behavior.
Even if you can get general trends right...inventory building versus reduction, unit increases versus ASP declines, etc. It only takes a one or two month delay to ruin numbers for this year and push any resulting gains/loses out in to next year.
By the time you put the impact of one-off events, of which there are always some in any given year, it is almost impossible to predict with any level of accuracy more than about 3 months ahead.
It seems analysts are not helping. If they lower and raise based on the dynamic nature of the economy, why should people pay them? I expect them to get facts out of nothing to have value for the subscriptions.
Blog Doing Math in FPGAs Tom Burke 7 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...