Predicting the Recovery

Brief note from the editor:

People often compliment me on my workload sharing techniques. For those who know me well, it will therefore come as no surprise that today’s article is written by a guest blogger.

Mike Cowan is a 45-year semiconductor industry veteran who worked for 36 years at IBM’s Microelectronics Division in East Fishkill, N.Y. Mike has developed the Cowan LRA Model for forecasting global semiconductor sales.

Mike Cowan’s key indicator of the health of the semiconductor industry is his “Momentum Indicator” or MI. The good news is that the MI has shown a dramatic improvement since its low point last December when actual sales data were almost 35% below the amount forecast in the previous month.

The rest I will leave to Mike’s explanation. He has an excellent set of slides the explain the model as well as providing a lot of useful current data and analysis about the semiconductor market that he will make available if you would like to contact him. His email is

mikedcowan(at)verizon.net.

Just mention that you read about it on SemiSerious, and he will get it to you right away.

- Don

The Cowan LRA Model For Forecasting Global S/C Sales – Overview Of Model And Latest Forecast Results

Model Overview

A semiconductor sales forecasting model has been developed to facilitate the determination of future global sales of the semiconductor industry. The Cowan LRA (Linear Regression Analysis) Model, which forecasts global semiconductor sales, is a mathematically based model that features statistical analysis of the past 25 years of historical, monthly global semiconductor sales numbers that are collected and published by the World Semiconductor Trade Statistics (WSTS) organization. It is a dynamic, mathematically pure view of near-term worldwide semiconductor sales looking forward over the next five quarters. The model is devoid of economic assumptions or emotional biases. It exploits linear regression analysis operating on the appropriately transformed actual monthly semiconductor sales numbers, thereby rendering the global semiconductor sales data highly linear and, therefore, very amenable to linear regression statistical analysis techniques. The numerical transformation of the past 25 years of monthly actual sales numbers — from 1984 through 2008 — that is invoked is not a complicated mathematical expression but very straight forward and makes sense physically, yielding extremely high linear regression correlation coefficients of 0.97 and greater. In exercising the forecasting model each month, a total of five distinct sets of linear regression parameters (of the format y = mx + b) are employed to calculate the resulting global semiconductor sales forecast estimates for each of the next five quarters associated with the model’s forecast horizon.

Cowan LRA Momentum Indicator

It is emphasized that each month’s actual global sales number released by WSTS is a lagging indicator because it is published a full month “after the fact.” The Cowan LRA Model, however, turns this lagging monthly sales result into a leading indicator by virtue of its near-term forecasting capability looking out over the next five quarters. This is the “beauty” of the model and, therefore, makes it dynamic in the sense that it can be run each month using the most recent actual global S/C sales number. Thus it can rigorously track the near-term sales forecast outlook of the global semiconductor industry on a real-time basis. Consequently, the model’s monthly sales forecast does not sit still but evolves each month because conditions change rapidly and unexpectedly in the semiconductor industry, and market forecasts are hard pressed to keep up with these changes. How can industry management be sure that a forecast issued two, three or more months ago is still relevant to what’s happening in today’s market?

Latest Sales Forecast Results

The latest global S/C sales forecast estimates, as derived via exercising the Cowan LRA Model, are based upon the recently-released (on Thursday, 6-04-09) April 2009 actual global S/C sales by the WSTS (note – includes, by the way, very minor upward revisions – i.e., increase in sales – for each of the previous three months of the year, namely, Jan., Feb., and Mar.).

The year 2009 sales forecast estimate kicked up strongly (by $8.52 billion) to $192.50 billion from last month’s sales forecast estimate of $183.99 billion with a corresponding improvement (increase) in the 2009 yr-o-yr sales growth forecast to minus 22.6 percent (from last month’s sales growth forecast estimate of minus 26.0 percent). It should be pointed out that this latest updated sales growth forecast estimate (minus 22.6 percent) is in good agreement with the latest forecast revisions just released by Gartner (minus 22.4 percent), the WSTS (minus 21.6 percent), and the SIA (minus 21.3 percent).

It should be highlighted that this month’s momentum indicator improved (increased) significantly to plus 19.6 percent. This is relatively good news when compared to the actual monthly momentum indicators from Oct08 through Jan09, that is -12%, -26%, -35%, and -31%, respectively. Historical tracking of this particular indicator over the past seven years reveals that the historical long-term trend of this indicator on a monthly basis correlates quite well with the significant collapse in industry sales during the time period just mentioned. Additionally the monthly trend shows a corresponding reversal to be in the making based upon this month’s strongly positive momentum indicator (a record high positive percent since the author has been running the model and tracking this particular indicator).

With this month’s momentum indicator, MI, moving into very high, positive territory bodes well relative to a possible relatively strong recovery in sales over the near-term. This monthly indicator bears continued watching over the coming months in order to monitor the trend of this indicator in order to verify that the present very strong uptick is highly suggestive that a “turning point” in the industry’s near term sales is real and will be sustained. However, the model’s predicted sales forecast estimate for next month (May 2009) is projected to be $14.686 billion which would represent a yr-o-yr monthly sales growth of minus 27.5 percent – not very encouraging relative to sustainability of a sales improvement trend if the model’s May09 sales forecast estimate is born out. Time will tell ==> stay tuned for next month’s update.

About The Author

Mike Cowan, the developer of the Cowan LRA Model, is a 45-year semiconductor industry veteran. He has a 36-year history at IBM’s Microelectronics Division in East Fishkill, N.Y., where he was involved in many facets of semiconductor development and manufacturing engineering, including both technical and management responsibilities. Over his last ten years at IBM, as a senior technical staff member, he was involved in strategy development and competitive analysis focused on the semiconductor industry, and has developed a number of top-down and bottom-up models to predict the dynamics of the semiconductor industry. After retiring from IBM in 2002 he became an independent semiconductor industry analyst providing his monthly forecasts to The Semiconductor Reporter website from 2002 to 2006 and to Future Horizons during 2007 and 2008.

Cowan earned both BS and MS degrees in physics at Wayne State University in Michigan, and an MS in electrical engineering at Syracuse University in New York.

For more information regarding the model or the results, please contact Mike directly at: mikedcowan(at)verizon.net.

Tags: Cowan, Global Semiconductor Forecast, Semiconductor Recovery
Posted in Industry News | No Comments »