Forecasting Is An Inexact Technology: Chip Analysts' Crystal Balls Are Clouded

By Roland Piquepaille

Predictions like "this specific market will increase from $50 million in 2002 to $31.48 billion in 2007" are common -- and most are plain wrong.

Today, Matthew Yi shared my skepticism by looking at some semiconductor industry's forecasts.

Consider what some of the leading firms are currently saying about 2003. They all believe next year's chip sales for the industry will improve from this year, but the growth rate ranges from 9 percent (IDC) to 26 percent (Semico Research Corp.). Gartner expects 12 percent growth; IC Insights, 15 percent; In-Stat MDR, 18 percent; VLSI Research Inc., 21.6 percent; and the Semiconductor Industry Association pegs the number at 19.8 percent.

To say the least, there is a wide range of variation. So let's look to the past to see if these research companies fared better.

In 1997, some expected the market to grow more than 10 percent, but it actually ended up around 4 percent. The following year, researchers were even more bullish, some forecasting as high as 23.5 percent growth. But the market actually shrank, down a little more than 8 percent. The year 2000 was another bullish year, but no one was near the actual growth of nearly 37 percent from the previous year.

Some forecasters say their job has become more difficult as the overall electronics industry has grown by leaps and bounds in the past decade. In the case of the chip industry, the emergence of contract manufacturing also has complicated the supply chain, which makes inventory tracking a more difficult task, industry researchers say.

When forecasts miss so badly, one can't help but wonder if industry analysts have a good handle on demand trends in the market place, said Michael Mahoney, high-tech portfolio manager at EGM Capital in San Francisco.

Even forecasters themselves admit they can be wrong.

Dan Hutcheson, who has been tracking the industry for more than two decades at VLSI Research Inc. in San Jose, doesn't hesitate to admit the accuracy of his annual forecasts has gone south in recent years. Between 1981 and 1994, his annual forecasts were off on average by about 5 percent, but in the past five years, he has been off by about 25 percent.

Still, many large companies pay more than $100,000 per year to read these reports -- and base their decisions to build a $1 billion factory upon them. Hey, if these analysts were wrong, they still can fire thousands of employees to please Wall Street.

The San Francisco Chronicle provides some interesting graphics about previous missed forecasts.

Source: Matthew Yi, San Francisco Chronicle, December 8, 2002