Background

Cluster development, however, should be derived from cluster growth. Economic growth can primarily be explained and measured by per capita income output (Marshall, 1920; Rostow, 1960; Hicks, 1946; Lewis, 1955). Likewise, usually the growth of industrial cluster can be evaluated by annual industrial output changes. This research applied historical quantitative forecasting methods to measure the growth of industrial clusters.

The historical forecasting of annual output in high-tech industries is useful for companies to prepare marketing strategies and perform production capacity planning and for financial institutions to make investment decisions (Chang, Lai, and Yu, 2003). This work attempts to forecast the annual output of semiconductor and computer industries of Hsinchu Science Industrial Park (HSIP) historically. HSIP provides a unique environment for accelerating technological innovation, nurturing new start-up firms, attracting investment, and generating economic growth (Potworowski, 2002). According to the historical forecasting analysis, industrial practitioners can understand market trends and customer’s needs and, in turn, modify their production strategies and replan their resources and capacities.

Taiwan is one of the world’s largest manufacturers of high-tech components and products. According to the World Economic Forum’s “2007-2008 Global Competitiveness Report,” Taiwan has again taken first place in the world in the “State of Cluster Development” index (see Appendix 1). HSIP was established in northwestern Taiwan as a focal point for high-tech R&D and production. Since the park’s beginnings, in 1980, the government has invested approximately US$3.67

billion in it. By the end of March 2007, 475 high-tech companies in six industries (integrated circuits, computers and peripherals, telecommunications, optoelectronics, precision machinery, and biotechnology) were situated within the 625-hectare park.

At the end of 2006, the park’s total paid-in capital exceeded US$35.4 billion, and more than 100 park companies were listed on Taiwan's main exchange, the Taiwan Stock Exchange, and on over-the-counter markets (Chen, 2007).

We can look to past research studies to observe the many historical forecasting methods that have been employed over the years, including quantitative and qualitative methods. Qualitative forecasting methods include the expert system and the Delphi method. Quantitative forecasting methods then include regression analysis, exponential smoothing, neural networks, and the Grey forecasting model (Chang et al., 2003).

Of the various forecasting models, the exponential smoothing model has been found to be one of the most effective. Since Brown (1959) began to use simple exponential smoothing to predict inventory demand, exponential smoothing models have been widely used in business and finance (Gardner, 1985). However, exponential smoothing methods are only a class of linear model, and, thus, it can only capture linear features of financial time series. Furthermore, as the smoothing constant decreases exponentially, the disadvantage of the exponential smoothing model is that it gives simplistic results that only use several previous values to forecast the future. The exponential smoothing model is, therefore, unable to find subtle nonlinear patterns in the financial time series data.

The Grey forecasting model has numerous applications. Hsu and Chen (2003) examines the precision of the Grey forecasting model applied to samples based on demand and sales in the global integrated circuit industry. Lin and Yang (2003) apply

of Taiwan’s opto-electronics industry from 2000 to 2005. Chang (2004) uses a grey forecasting model GM (1, 1) to improve the estimation of systematic risk of the classical capital asset pricing model.

In order to improve forecast accuracy, many researchers have modified the GM (1, 1) model. Liang, Zhao, Chang, and Liang (2001) utilized an improved grey model GM (1, 1) combined with a statistical method developed to evaluate the durability of concrete bridges due to carbonation damage. Yao, Chi, and Chen (2003) presented an improved Grey-based prediction algorithm to forecast a very short-term electric power demand for the demand-control of electricity. Chang, Lai, and Yu (2005) constructed a rolling Grey forecasting model (RGM) to predict Taiwan’s annual semiconductor production. Tien’s (2005) grey dynamic model DGDM (1, 1, 1) was first combined with the Grey-Markov chain forecasting model to predict the time for which the deviation is over the limit of tolerance deviation. Lin and Lee (2007) proposed a novel forecasting model termed MFGMn (1, 1) and modified the algorithm of the grey forecasting model to enhance the tendency catching ability.

Wang (2004) provided empirical evidence using grey theory and fuzzy time series, which do not require a large sample and long past time series. Yao et al. (2003) then presented an improved Grey-based prediction algorithm to forecast a very short-term electric power demand for the demand-control of electricity.

In the current research, we mainly apply quantitative historically forecasting methods to predict the annual output of computer and semiconductor industries in HSIP. First, we use an exponential smoothing method to historical forecast the annual output. Then, we apply the historically grey forecasting model to predict in the same regard. This research has also adopted a novel high-precision historical forecasting model, the Grey-Markov model, to enhance the prediction accuracy. Finally, we compare the prophecy accurateness among these three historical forecasting methods.

Again, we applied these historically quantitative forecasting methods to try to measure the growth of industrial clusters by observing and analyzing the changes of related industrial output at HSIP.

2.2 Taiwan’s Semiconductor and Computer Industries in Hsinchu Science Park