Technological Forecasting & Social Change 73 (2006) 422 – 435

Analysis of the supply and demand in the TFT–LCD market

James T. Lina, Fu-Kwun Wangb,*, Shin-Lian Loa, Wan-Ting Hsua, Yen-Tai Wangc
a
Department of Industrial Engineering and Engineering Management, National Tsing-Hua University, Kuang Fu Road,
Hsinchu 300, Taiwan, ROC
b
Department of Management, National Taiwan University of Science and Technology, No. 43 Keelung Road,
Sec.4, Taipei, 106 Taiwan, ROC
c
Marketing Division, Desktop Display Business Unit, AU Optronics Corporation, Science-Based Industrial Park,
Hsinchu 300, Taiwan, ROC

Received 26 May 2004; received in revised form 13 September 2004; accepted 14 September 2004

Abstract

A methodology for the analysis of the supply and demand in the Thin-Film Transitor (TFT)-liquid crystal
display (LCD) market in 2004 is proposed. The quarterly history supply and demand data are collected from 63
factories in Taiwan, Korea, Japan, and China during 2000–2003. This method takes into the account supply,
demand, and differences between supply/demand. For the supply, a heuristic approach is used to forecast the future
supply. For the demand, a transfer function model is used to forecast the future demand. The difference analysis of
the supply and demand shows that it can predict whether or not there appears to be a shortage in the market of
2004. In addition, three important managerial implications such as pricing strategy, product-mix decision, and
customer’s priority are discussed in this paper.
D 2004 Elsevier Inc. All rights reserved.

Keywords: Supply; Demand; TFT-LCD; Transfer function model

1. Introduction

Display became a critical interface for connecting the information during the 21st century. The Flat
Panel Display (FPD) plays a major role in exploring the advancement of modern IT devices. Chung et al.

* Corresponding author.
E-mail address: fukwun@mail.ntust.edu.tw (F.-K. Wang).

0040-1625/$ - see front matter D 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.techfore.2004.09.006
 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 423

[1] discusses the megatrend of FPD technology. Low temperature polycrystalline silicon (LTPS)-based
display device first became a major trend. The active LTPS display, where advanced circuits are
integrated on the panel, may continuously adopt liquid crystal. Secondly, the trend may include the
potential of adopting organic light-emitting device (OLED) as a display component replacing the current
liquid crystal and the fluorescent backlight system. Finally, the trend brings a flexible plastic substrate for
the TFT back plane and a color filter to make a flexible display. Mentley [2] concluded that there are no
limits to creativity of the display developers or to the tools available for discovery and development. The
challenge is to find the right combination of display performance and manufacturing cost to make FPDs
attractive and affordable to the masses of people, which in turn will enrich their lives through technology.
The current FPD market has been dominated by Thin-Film Transitor (TFT), which has Amorphous
Silicon (a-Si) to run an active matrix liquid crystal display (LCD). The large-size TFT-LCD has many
applications such as a Notebook PC, a desktop LCD monitor, a mobile, and an LCD-TV. Mentley [2]
also reported that the a-Si TFT-LCD would dominate FPD technology until 2010.
The motivation for this research stems primarily from the following issues associated with forecasting
in the TFT-LCD market. First, the supply of TFT-LCD is the current capacity for the manufacturer of the
TFT-LCD, and the ability and rate of manufacturer to add additional capacity if new technology can be
implemented. Second, the demand for TFT-LCD is a prediction of how much TFT-LCD is purchased.
The gap between supply and demand is whether or not there is excess in either demand or supply.
Finally, depending on the degree of the gap between supply and demand, either the supply or capacity
will increase the maximum rate possible or the proportion of the maximum rate required will satisfy
future predictions of demand. A shortage will continue for multiple periods if either predictions or
assumptions of the demand of the possible market underestimate the actual demand. On the other hand, a
shortage may exist for multiple periods because future demand exceeds the ability to bring new capacity
for satisfying demands immediately. Shortages in capacity allow manufacturers to modify their strategy
and to obtain maximum profits.
Zhao et al. [3] evaluated the impact of forecasting models and early order commitment in a supply
chain with one capacitated manufacturer and four retailers under demand uncertainty. They noticed that a
proper forecasting model would enhance the effectiveness of early order commitment. Chan et al. [4]
proposed a framework of a central coordination system, which is equipped with a multicriteria genetic
optimization feature to assist collaboration activities. The optimization results indicate the influences of
different sets of weightings on the demand allocation results. In this study, we examine the supply and
demand in the TFT-LCD market. Based on the different analysis of the supply and demand, we can
predict whether or not there appears to be a shortage in the market of 2004. This article is divided into
the following sections. The development of the TFT-LCD industry in Japan, Korea, and Taiwan is
reviewed in Section 2, followed by the forecasting model of the supply/demand, and lastly, the results.
Here, the forecasting of the supply is used by a heuristic approach, and a transfer function model is used
as the forecasting of the demand. The difference analysis between supply and demand and managerial
implications are presented in Section 4, and the conclusions are made in the final section.

2. The development of the TFT-LCD industry

The development of LCD began in 1964. Kawamoto [5] provided the details of the history where
development spans the world’s major industrial centers: the USA, Europe, and Japan. Each industrial
424 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

center contributed to their particular strengths: in America, it was the quickness of forming new ideas
and demonstrating their feasibility; in Europe, it was the fundamental science and synthesis of basic
materials; and in Japan, it was the process of perfecting implementation and moving it to the production
line. In 1992, Japan companies such as Toshiba, Matsushita, and NEC started the large size
manufacturing of the TFT-LCD. Korea companies such as LG Electronics and Samsung Electron Device
(formerly Samsung Display Device) have carried out researches on TFT-LCD since 1985 while doing
business in simplex-matrix LCD. A large-scale project and a gigantic amount of money were invested in
1995 in order to make production of TFT-LCD larger than 12.1 in. The Korean market share of TFT-
LCD is growing continuously and reached a 37.5% in 2000 and then exceeded to 40% in 2001. Jang et
al. [6] provided the details of the development of the TFT-LCD in Korea. Because of the Asian financial
crisis in 1997, Japan companies decided to transfer the manufacturing technology to Taiwan. In 1998,
seven Taiwan companies started the TFT-LCD manufacturing with a fab 3.5 generation. Recently,
Taiwan companies such as AUO, CMO, CPT, Hannstar, and Quanta Display invested more than US$ 3
billion into the development of manufacturing technologies such as process automation and optimal
control for large area lithography on the panel glass. In 2006, the production value of the display
industry in Taiwan will be more than US$ 40 billion. At the end of 2001, the market of the LCD
monitors exceeded US$ 20 billion. Currently, Korea, Taiwan, and Japan are the major countries for
making the TFT-LCD production.

3. Forecasting of the supply and demand in the TFT-LCD market

Fig. 1 depicts the framework of the TFT-LCD market forecasting. The quarterly data collection period
is during 2000–2003. These historical data will be used to forecast the future supply and demand in
2004. The first part is about the supply of the glass surface area. The second part is about the demand of

Fig. 1. The framework of the TFT-LCD market.

 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 425

the glass surface area. The third part is about the difference analysis of supply and demand, discussed in
the next section.

3.1. The supply model

In order to obtain the supply model, several variables such as capacity, yield rate, utilization rate,
and key components are considered. Also, key components such as glass, color filter, driven IC,
partial-light board, and backlight module affect the supply in the TFT-LCD market. Here, the
supply is based on the module shipment area. Data collection includes the fab generations 1–6,
which has 63 factories in Taiwan, Korea, Japan, and China during 2000–2003. These TFT-LCD
companies are:

Taiwan—AUO, CMO, CPT, Hannstar, and Quanta Display;

Korea—L.G., Philips, and Samsung;
Japan—ADI, Fujitsu, Hitachi, NEC, Sharp, Tottori Sanyo, and TMD;
China—BOE-Hydis.

To calculate the module shipment area for the period during 2000–2003, four major variables such as
efficiency ratio, capacity, utilization rate, and yield rate are used. Fig. 2 depicts how to calculate the
efficiency ratio for the glass substrate dimensions (11001250 mm) by the fab generation 5. This glass
can be divided into nine small glasses, where each glass has a diagonal of 19 in. The efficiency ratio is
defined as the cutting area divided by the substrate area. The capacity of each factory depends on the
fab generation used. According to the data from Display Search (www.displaysearch.com), the
utilization rate of the factory in 2000–2002 was about 80%, but most factories may have more than
90% utilization rate after 2002. The yield rate depends on three major processes—Array, Cell, and
Module. The yield rates for these three processes are about 90%, 90%, and 95%, respectively. Thus, the
overall yield rate is about 80–90%. However, it will take more efforts to reach the predicted yield rate

Fig. 2. The efficiency ratio of an example panel.

426 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

for a larger size of the glass. Therefore, we can define the total module shipment area (mm2) for the
history quarterly supply as:

X
63
Total module shipment are a per quarter ¼ li  wi  ei  ci  ui  yi ð1Þ
i¼1

where l i =the length of the glass on the ith factory’s production line, w i =the width of the glass on the ith
factory’s production line, e i =the efficiency rate of the glass on the ith factory’s production line, c i =the
capacity of the ith factory’s production line, u i =the utilization rate of the ith factory’s production line,
and y i =the yield rate of the ith factory’s production line.
For the future supply of the module shipment area, the calculation is given in Fig. 3. It is similar to the
above Eq. (1). However, the total module shipment area is based on whether there is enough supply of
the key components. Thus, the total module shipment area (mm2) for the future quarterly supply is
defined as:

X
63
Total module shipment are a per quarter ¼ li  wi  ei  ci  ui  yi  ð1  si Þ ð2Þ
i¼1

where s i =the key components shortage of the ith factory’s production line.
In this part, we used a heuristic approach to forecast the future supply. Four scenarios are assumed to
calculate the future supply in 2004. Scenario 1 is considered to not lack any key components. Scenario 2

Fig. 3. The calculation of the panel for the future supply.

 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 427

is considered to lack the key components of 2004, which has an 18% shortage during the first quarter, a
15% shortage during the second quarter, a 12% shortage during the third quarter, and a 10% shortage
during the fourth quarter. Scenario 3 is considered to lack key components of 2004, which has a 10%
shortage during the first quarter, an 8% shortage during the second quarter, a 7% shortage during the
third quarter, and a 5% shortage during the fourth quarter. Scenario 4 is considered to lack key
components of 2004, which has a 10% shortage during the first quarter, a 15% shortage during the
second quarter, a 20% shortage during the third quarter, and a 25% shortage during the fourth quarter.
According to the discussion above, the total module shipment area for the history period (2000–2003)
and the future period (2004) is shown in Fig. 4, where the future supply in 2004 is based on the four
different scenarios.

3.2. The demand model

Demand modeling is a complex problem due to the variety customer types and the relatively limited
amount of information that can be used for model parameter estimation. Here, a transfer function model,
which combines a time series and a multiple linear regression equation, is used to forecast the demand in
2004. This demand model is based on the shipment area for the time series model and the leading
indicators such as consumer index, industrial production index, GNP, economic growth rate in the USA,
Japan, Germany, and England for the multiple linear regression equation. Data collection for the
shipment is from the Display Search report and the IDC report. The leading indicators are from the
websites of http://stats.bls.gov/, http://www.fedstats.gov/, http://www.statistics.gov.uk/, http://www.
destatis.de/e_home.htm, and http://www.stat.go.jp/.
According to the varieties of the TFT-LCD products such as LCD monitor, Notebook, LCD TV, and
LCD Audio–Video, the shipment area for each product must be determined. For example, the resolution
of the 17 in. TFT-LCD Monitor called SXGA (12801024 mm), with the ratio of the length to the width

Fig. 4. The supply forecasting of the TFT-LCD market.

428 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

as 5:4, the area as 91011.5 mm2, and obtained by [17(in.)52/(52+42)0.525.4(mm/in.)][17(in.)42/

(52+42)0.525.4(mm/in.)]. Another example is the resolution of the 17 in. TFT-LCD Monitor called XGA
Wide (1280768 mm), with the ratio of the length to width as 5:3, the area as 82291.2 mm2, and
obtained by [17(in.)52/(52+32)0.525.4(mm/in.)][17(in.)32/(52+32)0.525.4(mm/in.)]. Thus, the demand
area of each product can be easily obtained. Accordingly, the total shipment area (mm2) for the history
quarterly demand is defined as:

X
4
Total module shipment are a per quarter ¼ li  wi  ri  qi ð3Þ
i¼1

where l i =the length of the ith product, w i =the width of the ith product, r i =the real area ratio of the ith
product which is defined as the cutting area divided by the active area, and q i =the shipment quantity of
the ith product.
For the future demand of the total shipment area in 2004, a multiple input Box–Jenkins model is used
in this study [7]. This model is a time series modeling process that describes a single dependent series as
a function of its own past values and the values of one or more independent input series. As with
univariate modeling, the purpose of multiple input modeling is to find the model that accounts for the
predictable portion of the dependent series. Such a model can then be used for both forecasting and
control. The model uses a weighted average of the past of both the input and output series to best fit the
data.

Y t;1 ¼ V0 þ V1 Yt1;1 þ V2 þ Yt2;1 þ : : : þ At þ W1 Xt;1 þ W2 Xt;2 þ : : : ð4Þ

Here, output series refers to the dependent series ( Y), which is a function of one or more
independent input series (X). The input series may either be stochastic (nondeterministic) or
intervention (deterministic). The transfer function accounts for the measurable relationship between
the output and each of the input series. The difference between that parts of the output series can be

Fig. 5. The demand forecasting of the TFT-LCD market.

 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 429

accounted for by the input series called the noise. The noise itself is stochastic and therefore may be
represented by an ARIMA model. The combined transfer function and noise model comprises the
multiple input Box–Jenkins model. The software Autobox [8] is used to determine the best-fit model.
Thus, we have:
 1
Y t ¼  0:49949  1011 þ 0:31589  109  Xt;1 þ 1 þ 0:644B2  At ð5Þ

where Y=total shipment area, X 1=CPI_A (consumer price index in the USA), A=noise term, and B is
called the backward shift operator, which is defined by BY t =Y t1. The R2 for the model is 0.984. Other
measures such as AIC value and BIC value are 384.561 and 406.519, respectively. The residual
analysis shows that the assumptions of the transfer function model are correct. According to the
discussion above, the total module shipment area for the history period (2000–2003) and the future
period (2004) is shown in Fig. 5. With respect to the first two quarters of 2004, using the model, we
have the mean absolute percent errors (MAPE) 2.13% and 3.52%, respectively. The model works quite
well.

4. The difference analysis between supply/demand and managerial implications

4.1. The difference analysis between supply/demand

According to the supply and demand analysis in Section 3, the difference analysis for four scenarios is
discussed in the following:

(1) Scenario 1: There is no supply shortage of the key components. Fig. 6 depicts the forecasting of the
total area in the supply and demand for the TFT-LCD market. In order to simply the analysis of the
relationship between the supply and demand, a new variable is defined as (supply–demand)/

Fig. 6. The supply and demand forecasting for Scenario 1.

430 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

Fig. 7. The difference of the supply/demand for Scenario 1.

demand. If this variable is greater than zero, it indicates that the supply is greater than the demand.
That is, there is no shortage in the market. On the other hand, if this variable is less than zero, it
indicates that the demand is greater than the supply. That is, there is shortage in the market. The
difference analysis between supply and demand is shown in Fig. 7. The results show that shortage
will not appear in the market after the end of the first quarter in 2004.
2) Scenario 2: There exists a supply shortage of the key components in 2004, i.e., Q1—18%, Q2—
15%, Q3—12%, and Q4—10%. Fig. 8 depicts the forecasting of the total area in the supply and

Fig. 8. The supply and demand forecasting for Scenario 2.

 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 431

Fig. 9. The difference of the supply/demand for Scenario 2.

demand for the TFT-LCD market. The difference analysis between supply and demand is shown in
Fig. 9. The results show that no shortage will appear in the market by the end of the third quarter in
2004.
3) Scenario 3: There exists a supply shortage of the key components in 2004, i.e., Q1—10%, Q2—8%,
Q3—7%, and Q4—5%. Fig. 10 depicts the forecasting of the total area in the supply and demand for
the TFT-LCD market. The difference analysis between supply and demand is shown in Fig. 11. The
results show that a shortage will not appear in the market by the end of the second quarter in 2004.
4) Scenario 4: There exists a supply shortage of the key components in 2004, i.e., Q1—10%,
Q2—15%, Q3—20%, and Q4—25%. Fig. 12 depicts the forecasting of the total area in the

Fig. 10. The supply and demand forecasting for Scenario 3.

432 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

Fig. 11. The difference of the supply/demand for Scenario 3.

supply and demand for the TFT-LCD market. The difference analysis between supply and
demand is shown in Fig. 13. The results show that a shortage will always exists in the market
in 2004.

According to the data from the AU Optronics, the supply shortage of the key components in the
first two quarters of 2004 is about 15–20%. They also estimate that the supply shortage of the key
components in the third and forth quarters of 2004 is about 10%. So, Scenario 2 is close to this
situation. The results show that no shortage will appear in the market by the end of the third quarter
in 2004.

Fig. 12. The supply and demand forecasting for Scenario 4.

 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 433

Fig. 13. The difference of the supply/demand for Scenario 4.

4.2. Managerial implications

The methodology proposed in this article can be utilized to evaluate whether or not the TFT-LCD
market is an equilibrium situation. The results show that the TFT-LCD market was not equilibrium
during 2000–2003. The same situation will apply for the future. The study performed in this work has
three important managerial implications such as pricing strategy, product-mix decision, and customer’s
priority.

(1) Pricing strategy: The TFT-LCD prices change can be defined as DP t =k(D t S t )+e t , where e t ~N(0,
r2t ). Here, the change in TFT-LCD price is assumed to be directly proportional to the amount of
excess demand in the market. The proportionality k depends on the length of the time unit, where
k=0 implies no adjustment, and k=l implies perfect adjustment. This equation suggests that the
direction of the TFT-LCD price change is an indication of the excess demand of the market. The
change in TFT-LCD prices will stabilize when demand equals supply. When the TFT-LCD price
change is positive, the manufacturer should focus on the improvement of the capacity and yield
rate. On the other hand, when the TFT-LCD price change is negative, the manufacturer should
focus down on the cost. Fig. 14 depicts the change of the average selling prices (ASP) during
2001–2003 and the difference analysis of the supply and demand during 2000–2003. The results
show that these two variables have a 0.6 relationship lag during one quarter.
(2) Product-mix decision: When the supply is greater than the demand, the company must face a
crucial competition in the market. In order to pursue the maximum profits, the company should
focus on the different market and technology improvements. It is necessary to evaluate the profit
for each product. Therefore, the product-mix decision must be developed for maximizing the
profits.
(3) Customer’s priority: the customer’s priority is different when the supply is greater than the demand
and when the demand is greater than the supply. For instance, when the demand was greater than
434 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435

Fig. 14. The difference of the supply/demand for the price change.

the supply in a particular product I, customer A purchased a larger quantity than customer B. In this
situation, customer A should be classified as higher priority than customer B. On the other hand,
when the supply is greater than the demand in two particular products I–II, customer A only
purchased product I, but customer B purchased both products. In this situation, customer-B should
be classified as higher priority.

5. Conclusions

A method is offered to forecast the supply and demand in the TFT-LCD market in 2004. From the
different analysis between the supply and demand, it will be easy to predict when a shortage situation
will appear in the market. Furthermore, three important managerial implications, pricing strategy,
product-mix decision, and customer’s priority, are obtained. The company can obtain the maximum
profits from the study of forecasting the supply and demand in the market.
Further research issue can be focused on the effects of forecasting each type of products for the TFT-
LCD in the market. In addition, consideration of the collaboration forecasting among the supplier,
manufacturer, distributor, and retailer will make more benefits in this TFT-LCD supply chain. One
referee suggested that the demand in the Chinese market could be effectively infinite. Thus, the research
issue can be focused on how to predict the TFT-LCD in the Chinese market.

Acknowledgments

The authors wish to gratefully acknowledge the support from the AU Optronics in Taiwan for
providing the opportunity for this study. In addition, the authors gratefully acknowledge the referees of
this paper and the editor in chief who helped clarify and improve the presentation.
 J.T. Lin et al. / Technological Forecasting & Social Change 73 (2006) 422–435 435

References

[1] K. Chung, M.P. Hong, C.W. Kim, I. Kang, Needs and solutions of future flat panel display for information technology
industry, Proc. IEEE-IEDM (2002) 385 – 388.
[2] D.E. Mentley, State of flat-panel display technology and future trends, Proc. I.E.E.E. 90 (4) (2002) 453 – 459.
[3] X. Zhao, J. Xie, R.S.M. Lau, Improving the supply chain performance: use of forecasting models versus early order
commitments, Int. J. Prod. Res. 39 (17) (2001) 3923 – 3939.
[4] F.T.S. Chan, S.H. Chung, S. Wadhwa, A heuristic methodology for order distribution in a demand driven collaborative
supply chain, Int. J. Prod. Res. 42 (1) (2004) 1 – 19.
[5] H. Kawamoto, The history of liquid-crystal displays, Proc. I.E.E.E. 90 (4) (2002) 460 – 500.
[6] J. Jang, S. Lim, M. Oh, Technology development and production of flat panel displays in Korea, Proc. I.E.E.E. 90 (4)
(2002) 501 – 513.
[7] G.P. Box, G.W. Jenkins, G.C. Reinsel, Time Series Analysis Forecasting and Control, Prentice-Hall, Englewood Cliffs,
NJ, 1994.
[8] Autobox, Version 5.0, Automatic Forecasting Systems, Hatboro, PA, 2002.

James T. Lin received his PhD degree in Industrial Engineering at Lehigh University. He is a Professor in the Department of
Industrial Engineering and Engineering Management at the National Tsing-Hua University (NTHU), Taiwan, Province of
China. His current research and teaching interests are in the general area of Supply Chain and Production Management. In
particular, he is interested in Simulation Modeling of Manufacturing Systems or AMHS (Automated Material Handling
System), Advanced Planning, and Scheduling.

Fu-Kwun Wang received his MSc and PhD degrees in Statistics and Industrial Engineering from Arizona State University,
USA. Currently, he is an associate professor in the Department of Industrial Management at the National Taiwan University of
Science and Technology, Taiwan. His primary research interests are in applied statistics, statistical quality control, and
reliability. He has been published in a variety of journals including IEEE Transactions on Semiconductor Manufacturing,
Journal of Quality Technology, Total Quality Management, Reliability Engineering and System Safety, Omega, International
Journal of Production Research and Production Planning and Control, etc.

Shin-Lian Lo is a graduate student in the department of Industrial Engineering and Engineering Management at the National
Tsing-Hua University, Taiwan. Her research interests include forecasting analysis and production management in semiconductor
manufacturing.

Wan-Ting Hsu is a graduate student in the department of Industrial Engineering and Engineering Management at the National
Tsing-Hua University, Taiwan. Her research interests include forecasting analysis and production management in semiconductor
manufacturing.

Yen-Tai Wang is a manager in the Marketing Division, Desktop Display Business Unit, AU Optronics, Taiwan.