Reconfiguring the innovation policy portfolios for Taiwan's SIP Mall industry

(1)

Technovation 27 (2007) 744–765

Reconﬁguring the innovation policy portfolios

for Taiwan’s SIP Mall industry

Chi-Yo Huang

a,

, Joseph Z. Shyu

b

, Gwo-Hshiung Tzeng

b,c

a_{Department of Industrial Education, National Taiwan Normal University, 162, He-Ping East Road, Section 1, Taipei 106, Taiwan} b_{Institute of Management of Technology, National Chiao-Tung University, 1001, Ta-Hsueh Road, Hsin-Chu 300, Taiwan}

c

Department of Business Administration, Kainan University, No. 1 Kainan Road, Luchu, Taoyuan 338, Taiwan

Abstract

Since the year 2000, silicon intellectual property (SIP), which can minimize the gap in ‘design productivity’ that exists with systems-on-chip (SOC), has become one of the most important factors in the development of integrated circuit (IC) products in the SOC era. Although SIP is very important for IC industry development, complicated business, technical as well as legal issues inside SIP transactions have hindered successful transactions and the integration of SIPs into SOCs. Thus, web-based SIP e-Commerce mechanisms, called SIP Malls, have emerged, aiming to resolve complex SIP issues. To maintain its leading position and competitiveness in the World’s IC industry, as well as the value added by SOC products in Taiwanese IC firms, the Taiwanese government has developed SIP Malls, using innovation policy tools. However, the Taiwanese SIP Mall industry remains immature. No existing Taiwanese SIP Malls generate a profit or account for a significant share of worldwide SIP transactions. This research will develop an analytical framework for defining an innovation policy portfolio that aims to develop Taiwan’s SIP Mall industry, so that it will enhance the value added of SIP Malls and, thus, the nation’s competitiveness in the SIP and IC industries, something which already has become one of the Taiwanese government’s major concerns. The industry innovation requirements (IIRs) are summarized using the Delphi method. Meanwhile, the major IIRs identified by Decision Making Trial and Evaluation Laboratory (DEMATEL) are introduced. After the IIRs are derived, the relationships between the IIRs and innovation policy tools are derived by Grey relational analysis (GRA). Then, the innovation policy tools are clustered, based upon the Grey grades derived by GRA. Finally, reconfigured innovation policy portfolios are presented for the Taiwanese government’s policy definition. The results demonstrate that developing an innovative policy portfolio that includes scientific, technical, educational, public enterprise, information, legal and regulatory, financial, and taxation policy tools will be the most necessary step towards developing Taiwan’s SIP Mall industry.

Keywords: Innovation policy; DEMATEL; Grey relational analysis; Integrated circuit (IC); Systems-on-chip (SOC); Intellectual property (IP); Design and reuse; SIP Mall; Semiconductor

1. Introduction

Capitalism is undergoing an epochal transformation from a mass production system, wherein the principal source of value has been human labor, to a new era of ‘innovation-mediated production’, wherein the principal component of value creation, productivity and economic growth is knowledge (Florida and Kenney, 1991). Knowl-edge has become a ﬁrm’s primary means of generat-ing proﬁts (Bekkers et al., 2002). In knowledge-based

industries, knowledge can be marketed to potential partners or clients separate from the products and services that are rooted in the application of speciﬁc knowledge (Borg, 2001). A key feature of the commercialization of knowledge is its acquisition and control as ‘intellectual property’ (IP) (Rappert and Webster, 1997). The produc-tion of new knowledge would be optimized by establishing strong IP rights that create incentives for generating knowledge (Adler, 2001).

Knowledge emerges in network relationships between an enterprise and its surrounding organizations, and can be protected legally as IP (Borg, 2001). IP, a legal term describing legally protected intellectual assets in their

www.elsevier.com/locate/technovation

Corresponding author. Tel.: +886 936 516698; fax: +886 3 5165185. E-mail address:georgeh168@gmail.com (C.-Y. Huang).

(2)

various available forms, is the subset of intellectual assets that is legally protected (Sullivan et al., 2002). Patents, trademarks, copyrights, trade secrets, and semiconductor mask works are ﬁve major forms of silicon IPs (SIPs) available in the integrated circuit (IC) industry (Sullivan et al., 2002). SIP has existed since the advent of the semiconductor industry. In the early years, IC suppliers— like Fairchild, Intel and Motorola—developed proprietary SIPs for their internal use. During the past decade, IC design productivity has failed to keep pace with Moore’s Law, which predicts that the number of electronic devices that can be fabricated on an IC chip will double every 18 months (Moore, 1979). Thus, a ‘design gap’ between IC design complexity increase and productivity increase has emerged (Semiconductor Industry Association, 2002). IC suppliers began looking for ways to narrow the gap by designing ICs with reusable SIP that tended to contain increasingly complex functionality (Ratford et al., 2003). As IC designs have become more complex, a large number of SIP products are being embedded into the designs. The SIP has become a key segment of the electronic design process, as it can reduce IC development costs, accelerate time-to-market, reduce time-to-volume, and increase end-product value. The nature of SIP, which can narrow the ‘design productivity gap’, has made SIP critical for the design and implementation of complex systems-on-chips (SOC), which have become the mainstream solution for realization of electronic system products since the year 2000.

Despite the optimistic future that exists for SIP, potential problems, including complex laws and regulations pertaining to SIP transactions, negotiations over SIP usage rights, technical support including SIP integrations into SOCs, maintenance, and SIP application engineering based upon current business models, have emerged as roadblocks to successful SIP business. Apparently, SIP transactions and integrations are not easy. Thus, an e-Commerce mechanism for SIP transactions, the SIP Mall—that aims to provide a well-established SIP database, to increase SIP transaction efﬁciency, and to provide SOC designers with well-veriﬁed reusable SIPs, design environments, and design services—can resolve the above-mentioned business, legal, and technical issues effectively, and be most helpful in accelerating SIP and IC industry growth.

The well-established SIP Malls ﬁll the vacancy that exists in the IC industry structure, resolve existing SIP transac-tion problems, and enable innovatransac-tions in IC. The upstream electronic system houses and the fabless IC design houses may develop synergies, by leveraging the SIPs and design services provided by SIP Malls and integrating the manufacturing capacities of IC foundries and assembly houses to roll out innovative SOC products at the most advanced process nodes. Thus, the well-established SIP Malls enhance the nation’s competitiveness in the IC industry.

Taiwan is the world’s largest semiconductor foundry and second largest fabless IC design provider. To keep the

leading position, competitiveness and the value added of Taiwan’s foundry, fabless, and thus the IC industry in the SOC era, the Taiwanese government supported the development of Taiwan’s SIP Mall industry, by leveraging several policy tools specifically targeting the SOC and SIP industry, including education; research and development (R&D) funding; taxation; and cooperation between R&D institutions, universities and private firms. For example, the Si-Soft project was formed in 2000, followed by the establishment of the Semiconductor Industry Group. Further, with a budget of 7.6 billion NTD, the Ministry of Economic Affairs, Ministry of Education and National Science Council coordinated the first phase of the National Si-Soft Project (Chang and Trappey, 2003) that was aimed at enhancing Taiwan’s capabilities in (1) innovative SOC product designs; (2) silicon SIP development; (3) electronic design automation (EDA) flow integration; (4) SIP Malls; and (5) SOC design services, spanning the years from 2002 to 2005. Even though the Taiwanese government has exerted considerable effort into developing the SIP and SIP Mall industries over the past 5 years, Taiwanese SIP Malls still are in the exploratory phase, while SIP sources are limited, and SIP transactions through SIP Malls remain few. How to develop the SIP Mall industry, and how to enhance the value-added nature of SIP Malls, and thus the competitiveness of Taiwan’s SIP and IC industries, already have become key concerns of the Taiwanese government.

Therefore, the purpose of this paper is to reconﬁgure Taiwan’s innovation policy portfolios for developing the SIP Mall industry, by analyzing industrial innovation requirements (IIRs) of the SIP Mall industry and propos-ing innovation policy portfolios based upon the IIRs derived.

Delphic Oracle’s Skills of Interpretation and Foresight (Delphi), Decision Making Trial and Evaluation Labora-tory (DEMATEL), Grey relational analysis (GRA), and cluster analysis all are used to build an analytical structural framework to derive the innovation requirements for developing Taiwan’s SIP Mall industry, reducing the over-complicated IIRs to the most significant IIRs, and then mapping the most significant IIRs to implement the innovation policy tools being proposed by Rothwell and Zegveld (1982). From these analytical results, the authors found that scientific and technical as well as education tools are the most important policy tools, while public enterprise, information, financial, taxation, and legal and regulatory tools also will be necessary to develop Taiwan’s SIP Mall industry.

The remainder of this paper is organized as follows. In Section 2, the concepts of innovation and innovation policy are introduced. In Section 3, an analytical framework and methods are proposed for constructing the IIRs and innovation policy portfolio deﬁnitions. These include ﬁrst collecting IIRs by Delphi, then deriving major IIRs by DEMATEL, and concluding with the key innovation policies developed by analyzing the relationships between IIRs and innovation policies using GRA. Finally, the

(3)

innovation policy tools are clustered as portfolios using cluster analysis. The background of SIP Mall industry development will be described in Section 4. Then, in Section 5, an empirical study follows, reconﬁguring the innovation policy portfolios for developing Taiwan’s SIP Mall industry. Discussions will be presented in Section 6. Section 7 will conclude the whole article with observations, conclusions and recommendations for further study. 2. Concepts of innovation, industrial innovation requirements and innovation policy

Researchers have successfully explored the deﬁnitions of innovation, innovation theories, the rationale of govern-ment interventions on innovation, innovation policy instruments, and the relationships between new technolo-gies, emerging markets, innovative services, and economic growth. In the following section, the related literature is reviewed.

Innovation is combinations of knowledge that result in new products, processes, input and output markets, or organizations (Sundbo, 2003) which include not only technical innovations, but also organizational and manage-rial innovations, new markets, new sources of supply, ﬁnancial innovations, and new combinations (Perlman and Heertje, 1991). Innovation is a critical factor in enhancing a

nation’s competitiveness. National governments have pursued planning in innovation policies to improve their nation’s growth. In a knowledge-based economy, innova-tion through the creainnova-tion, diffusion and use of knowledge has become a key driver of economic growth (Carayannis et al., 2006).

Industrial innovation includes technical design, manu-facturing, management and commercial activities involved in the marketing of a new (or improved) product or the ﬁrst commercial use of a new (or improved) process or equipment (Freeman, 1982). The factors required for industrial innovation or IIRs include technical knowledge, manpower, market information, management skills, ﬁnan-cial resources, R&D environments, a domestic market, and an international market (Fig. 1) (Rothwell and Zegveld, 1982). Industrial innovation, indeed, can increase overall economic development (Rothwell and Zegveld, 1982).

Most economic functions in a modern society are best fulfilled by the market mechanism and capitalist firms. However, sometimes there are reasons to complement—or correct—the market and its capitalist actors through public intervention, in such areas as law, education, environment, infrastructure, research, social security, and income dis-tribution. In some of these fields, there is no market mechanism at all, and the functions are fulfilled through use of other mechanisms, like regulation (Edquist, 2001).

(4)

Thus, government intervention may be needed at certain specific stages of the innovation process. For centuries, governments have pursued policies to improve the inno-vative performance of domestic industries, and to en-hance the economic returns to these domestic firms and their citizens from indigenously developed technologies (Mowery, 1995). The role of government in innovation is not monolithic. From supporting basic research to building infrastructures and establishing regulations, government policy-makers can define industries and affect the fortunes of individual firms (Hung and Chu, 2006). Meanwhile, governments exert a strong influence on the innovation process, via the financing and steering of public organiza-tions that are directly involved in knowledge generation and diffusion (like universities and public labs), and through the provision of financial and regulatory incentives (Carayannis et al., 2006). Taiwan’s Hsin-Chu science-based industrial park and Bangalore’s software cluster in India (Wonglimpiyarat, 2006) are typical examples of the success of governmental interventions in innovation. To summar-ize, a national government may want to play a role in the process of innovation, because of: (1) the ‘public’ nature of the knowledge that underpins innovation; (2) the uncer-tainty that often plagues the process of innovation; (3) the need for certain kinds of complementary assets; (4) the nature of certain technologies; and (5) plain politics (Afuah, 1998).

Edquist and Hommen (1999) deﬁned science, technol-ogy, and innovation policy (in the narrow sense) as speciﬁc parts of what could be labeled more broadly as ‘innovation policy.’ Science policy is the most supply-side-oriented of

the policies, and the least direct. Technology policy is the most difficult to define, because technological research varies significantly within the continuum, from relatively mono-disciplinary scientific research to multidisciplinary commercial innovation. However, innovation policy that is oriented toward appropriate new product ideas, produc-tion processes, and marketing concepts can produce competitive advantages (Edquist and Hommen, 1999).

The search for appropriate policy tools is not easy. Macro measures are not effective; thus, proposals like a general R&D tax credit are pointless. Policies must be designed to influence particular economic sectors and activities. In this regard, the key policy problem is to augment or redesign institutions, rather than to achieve particular resource allocations (Shyu, 1999). Government intervention can be in the form of financing R&D; acting as a lead user; providing complementary assets; regulating a firm’s activities; educating the workforce; maintaining macroeconomic funda-mentals; and maintaining political stability in order to attract investment in the innovation (Afuah, 1998). A list of possible innovation policies provided byRothwell and Zegveld (1982)

is categorized into 12 groups and summarized in Table 1. These groups represent the policies that affect the innovation environment and the technological systems. The aims and governmental expectations of public policies towards innova-tion are many and varied, as are the policies themselves and the tools designed to meet policy aims. Innovation policy tools generally can be grouped under three main categories: (1) Supply: provision of ﬁnancial, manpower and technical assistance, including the establishment of a scientiﬁc and technological infrastructure; (2) Demand: central and local

Table 1

Classiﬁcation of government policy tools

Policy tool Examples

Supply side

Public enterprise Innovation by publicity owned industries, setting up of new industries, pioneering use of new techniques by public corporations, participation in private enterprise

Scientiﬁc and technical Research laboratories, support for research association, learned societies, professional associations, research grants

Education General education, universities, technical education, apprenticeship schemes, continuing and further education, retraining

Information Information networks and centers, libraries, advisory and consultancy services, databases, liaison services Environmental side

Financial Grant loans subsidies, ﬁnancial sharing arrangements, provision of equipment, buildings or services, loan guarantees, export credits

Taxation Company, personal, indirect and payroll taxation, tax allowances

Legal and regulatory Patents, environmental and health regulation. inspectorates, monopoly regulations

Political Planning, regional policies, honor or awards for innovation, encouragement of mergers of joint consortia. public consultation

Demand side

Procurement Central or local government purchases and contracts, public corporations R&D contracts, prototype purchases Public services Purchases, maintenance, supervision and innovation in health service, public building, construction, transport,

telecommunications

Commercial Trade agreements, tariffs, currency regulations Overseas agent Defense sales organizations

(5)

government purchases and contracts, notably for innovative products, processes and services; and (3) Environment: taxation policy, patent policy and regulations (e.g. economic, worker health and safety, and environment) that comprise those measures that establish the legal and ﬁscal framework in which industry operates.

As stated by Salmenkaita and Salo (2002), there are no straightforward answers to the questions: what elements should an innovation policy include, and how should such policies be implemented? Thus, the current authors have chosen to focus on this topic, so as to define a quantitative innovation policy portfolio definition procedure, and then verified this procedure by analyzing the SIP Mall industry, using a case study from Taiwan’s SIP Mall industry. 3. Analytical framework and methods for innovation policy portfolio definition

The analytical framework for defining innovation policy portfolios is initiated by collecting the IIRs needed to develop a nation’s SIP Mall industry, using the Delphi method. Since any IIRs derived by Delphi may impact one another, major IIRs were identified using DEMATEL. Finally, GRA was applied to determine the correlation between the major IIRs and the innovation policy tools. Based upon the GRA results, the innovation policy portfolio was derived. In summary, this evaluation frame-work (Fig. 2) consists of four main phases: (1) establishing IIRs using the Delphi method; (2) simplifying the IIRs using DEMATEL; (3) correlating the IIRs and policy tools and, thus, ranking the priorities of the innovation policy tools with GRA; and finally, (4) deciding the innovation policy portfolios, by means of cluster analysis.

3.1. Delphi method

The Delphi method originated in a series of studies conducted by the RAND Corporation in the 1950s (Jones and Hunter, 1995). The objective was to develop a technique to obtain the most reliable consensus from a group of experts (Dalkey and Helmer, 1963). While researchers have developed variations of the method since its introduction, Linstone and Turoff (1975) captured its common characteristics in the following description: Delphi may be characterized as a method for structuring a group communication process; so the process is effective

in allowing a group of individuals, as a whole, to deal with a complex problem. To accomplish this ‘structured communication,’ certain aspects should be provided: some feedback of individual contributions of information and knowledge; some assessment of the group judgment or viewpoint; some opportunity for individuals to revise their views; and some degree of anonymity for individual responses (Linstone and Turoff, 1975). The Delphi technique enables a large group of experts to be surveyed cheaply, usually by mail using a self-administered ques-tionnaire (although computer communications also have been used), with few geographical limitations on the sample. Specific situations have included a round in which the participants meet to discuss the process and resolve any uncertainties or ambiguities in the wording of the ques-tionnaire (Jones and Hunter, 1995). The Delphi method proceeds in a series of communication rounds, as follows: Round 1: Either the relevant individuals are invited to provide opinions on a specific matter, based upon their knowledge and experience, or the team undertaking the Delphi expresses opinions on a specific matter and selects suitable experts to participate in subsequent questionnaire rounds; these opinions are grouped together under a limited number of headings, and statements are drafted for circulation to all participants through a questionnaire (Jones and Hunter, 1995). Round 2: Participants rank their agreement with each statement in the questionnaire; the rankings then are summarized and included in a repeat version of the questionnaire (Jones and Hunter, 1995).

Round 3: Participants re-rank their agreement with each statement in the questionnaire, and have the opportu-nity to change their score, in view of the group’s response; the re-rankings are summarized and assessed for their degree of consensus: if an acceptable degree of consensus is obtained, the process may cease, with the ﬁnal results then fed back to the participants; if not, this third round is repeated (Jones and Hunter, 1995).

3.2. DEMATEL method

The DEMATEL method was developed by the Battelle Geneva Institute: (1) to analyze complex ‘world problems’ dealing mainly with interactive man-model techniques; and

(6)

(2) to evaluate qualitative and factor-linked aspects of societal problems (Gabus and Fontela, 1972). The applic-ability of the method is widespread, ranging from industrial planning and decision-making to urban planning and design, regional environmental assessment, analysis of world problems, and so forth. To apply the DEMATEL method smoothly, the authors reﬁned the deﬁnitions by

Hori and Shimizu (1999) and Chiu et al. (2006), and produced the essential deﬁnitions indicated below.

The DEMATEL method is based upon graph theory, enabling us to plan and solve problems visually, so that we may divide multiple criteria into a cause and effect group, in order to better understand causal relationships. Directed graphs (also called digraphs) are more useful than directionless graphs, because digraphs will demonstrate the directed relationships of sub-systems. A digraph typically represents a communication network, or a domination relationship between individuals, etc. Suppose a system contains a set of elements, S ¼ fs1; s2; . . . ; sng, and particular pair-wise relationships are determined for modeling, with respect to a mathematical relationship, MR. Next, portray the relationship MR as a direct-relation matrix that is indexed equally in both dimensions by elements from the set S. Then, extract the case for which the number 0 appears in the cell (i, j), if the entry is a positive integral that has the meaning of:

the ordered pair (si, sj) is in the relationship MR;

it has the kind of relationship regarding that element such that sicauses element sj.

The digraph portrays a contextual relationship between the elements of the system, in which a numeral represents the strength of inﬂuence (Fig. 3). The elements s1, s2, s3and

s4represent the factors that have relationships in Fig. 9.

The number between factors is inﬂuence or inﬂuenced degree. For example, an arrow from s1to s2represents the

fact that s1 inﬂuences s2and its inﬂuenced degree is two.

The DEMATEL method can convert the relationship between the causes and effects of criteria into an intelligible structural model of the system (Chiu et al., 2006).

Deﬁnition 1. The pair-wise comparison scale may be designated as 11 levels, where the scores 0, 1, 2, y, 10

represent the range from ‘no inﬂuence’ to ‘very high inﬂuence.’

Definition 2. The initial direct relation/influence matrix A is an n n matrix obtained by pair-wise comparisons, in terms of influences and directions between the IIRs, in which aij is denoted as the degree to which the ith IIR

affects the jth IIR.

A ¼ a11 a12 a1n a21 a22 a2n .. . .. . .. . .. . an1 an2 ann 2 6 6 6 6 4 3 7 7 7 7 5.

Deﬁnition 3. The normalized direct relation/inﬂuence matrix N can be obtained through Eqs. (1) and (2), in which all principal diagonal elements are equal to zero.

N ¼ zA, (1) where z ¼ 1 , max 1pipn Xn j¼1 aij. (2)

In this case, N is called the normalized matrix. Since lim

k!1N

k_{¼ ½0.}

Deﬁnition 4. Then, the total relationship matrix T can be obtained using Eq. (3), where I stands for the identity matrix.

T ¼ N þ N2þ. . . þ Nk¼N ðI N Þ1, (3)

where k-Nand T is a total inﬂuence-related matrix; N is a direct inﬂuence matrix and N ¼ ½xijnn; lim

k!1ðN

2_{þ þ} NkÞ stands for a indirect inﬂuence matrix and 0pxijo1. So, lim

k!1N

k_{¼ ½0. Detailed calculation procedures are} illustrated in Appendix A.

The (i, j) element tij of matrix T denotes the direct and

indirect inﬂuences of factor i on factor j.

Deﬁnition 5. The row and column sums are separately denoted as r and c within the total-relation matrix T through Eqs. (4)–(6). T ¼ ½tij; i; j 2 f1; 2; . . . ; ng, (4) r ¼ ½rin1¼ Xn j¼1 tij " # n1 , (5) c ¼ ½cj1n¼ Xn i¼1 tij " # 1n , (6)

where the r and c vectors denote the sums of the rows and columns, respectively. s2 2 3 1 3 s1 s4 s3

(7)

Deﬁnition 6. Suppose ridenotes the row sum of the ith row

of matrix T. Then, ri is the sum of the inﬂuences

dispatching from factor i to the other factors, both directly and indirectly. Suppose that cjdenotes the column sum of

the jth column of matrix T. Then, cj is the sum of the

inﬂuences that factor i is receiving from the other factors. Furthermore, when i ¼ j (i.e., the sum of the row sum and the column sum (ri+cj) represents the index representing

the strength of the inﬂuence, both dispatching and receiving), (ri+cj) is the degree of the central role that

factor i plays in the problem. If (ricj) is positive, then

factor i primarily is dispatching inﬂuence upon the other factors; and if (ricj) is negative, then factor i primarily is

receiving inﬂuence from other factors (Tamura et al., 2002).

3.3. GRA method

SinceDeng (1982)proposed Grey theory, related models have been developed and applied to MCDM problems. Similar to fuzzy set theory, Grey theory is a feasible mathematical means that can be used to deal with systems analysis characterized by inadequate information. Fields covered by the Grey theory include systems analysis, data processing, modeling, prediction, decision-making, and control engineering (Deng, 1985, 1988, 1989; Tzeng and Tasur, 1994). In this section, we briefly review some relevant definitions and the calculation process for the Grey Relation Model. This research modified the

deﬁni-tions by Chiou and Tzeng (2001) and produced the

deﬁnitions indicated below.

GRA is used to determine the relationship between two sequences of stochastic data in a Grey system. The procedure bears some similarity to pattern recognition technology. One sequence of data is called the ‘reference pattern’ or ‘reference sequence,’ and the correlation between the other sequence and the reference sequence is to be identiﬁed (Deng, 1986;Tzeng and Tasur, 1994;Mon et al., 1995;Wu et al., 1996).

Definition 7. The relationship scale also may be designated into 11 levels, where the scores 0, 1, 2, y, 10 represent the range from ‘no relationship’ to ‘very high relationship’ between the specified IIR and the innovation policy. Definition 8. The initial relationship matrix G is an m n matrix, where there are m innovation policy tools and n IIRs, obtained by surveying the relationships, where gkiis

denoted as the relationship between the kth IIR and the ith innovation policy tool.

G ¼ g11 g1i g1n .. . . . . .. . . . . .. . g_k1 g_ki g_kn .. . . . . .. . . . . .. . g_m1 g_mi g_mn 2 6 6 6 6 6 6 6 6 4 3 7 7 7 7 7 7 7 7 5 .

Deﬁnition 9. The normalized relationship matrixG can be obtained through Eqs. (7) and (8).

p_i¼1 max 1pkpmgki, X ¼ x11 x1i x1n .. . . . . .. . . . . .. . xk1 xki xkn .. . . . . .. . . . . .. . xm1 xmi xmn 2 6 6 6 6 6 6 6 6 4 3 7 7 7 7 7 7 7 7 5 , (7) Xi¼piGi. (8)

Deﬁnition 10. Letx0be the reference pattern with n entries

(i.e. dependent variable): x0¼ ðx0ð1Þ; x0ð2Þ;. . . ; x0ðnÞÞ and xi, the matrix containing the normalized mapping

informa-tion of each innovainforma-tion policy tool to the IIRs, be one of the m patterns with n entries to be compared with the x0, where

xi is written as xi¼ ðxið1Þ; xið2Þ;. . . ; xiðnÞÞ, 1pipm: The sequence xigenerally expresses the inﬂuencing factor of x0.

Deﬁnition 11. Let X be a normalized factor set of grey relations, x02X the referential sequence, and xi2X the comparative sequence; with x0(k) and xi(k) representing the

numerals at point k for x0 and xi, respectively. If

gðx0ðkÞ; xiðkÞÞ and gðx0; xiÞ are real numbers, and satisfy the following four grey axioms, then call gðx0ðkÞ; xiðkÞÞ the grey relation coefﬁcient, and the grade of the grey relation gðx0; xiÞis the average value of gðx0ðkÞ; xiðkÞÞ.

1. Norm Interval

0ogðx0ðkÞ; xiðkÞÞp1; 8k; gðx0; xiÞ ¼1 iff x0¼xi; gðx0; xiÞ ¼0 iff x0; xi2f;

where f is an empty set. 2. Duality symmetric

x; y 2 X ) gðx; yÞ ¼ gðy; xÞ iff X ¼ fx; yg. 3. Wholeness

gðxi; xjÞ a often

gðxj; xiÞiff X ¼ fxiji ¼ 0; 1; . . . ; ng; n42. 4. Approachability

gðx0ðkÞ; xiðkÞÞ decreases when jðx0ðkÞ xiðkÞÞj increases.

Deng also proposed a mathematical equation for the grey relation coefﬁcient, as follows:

gðx0ðkÞ; xiðkÞÞ ¼ min 8i min8k ðx0ðkÞ xiðkÞÞ þ z max 8i max8k ðx0ðkÞ xiðkÞÞ ðx0ðkÞ xiðkÞÞ þ z max 8i max8k ðx0ðkÞ xiðkÞÞ , ð9Þ

(8)

where z is the distinguished coefﬁcient ðz 2 ½0; 1Þ. Gen-erally, we pick z ¼ 0.5. Detailed explanations of this equation are provided in Appendix B.

Deﬁnition 12. If gðx0; xiÞ satisﬁes the four grey relation axioms, then g is called the Grey relational map.

Definition 13. If G is the entirety of the grey relational map, gAG satisfies the four axioms of the grey relation, and X is the factor set of the grey relation, then (X, G) will be called the grey relational space, while g is the specific map for I0_. Definition 14. Let (X, G) be the grey relational space, and if gðx0; xjÞ; gðx0; xpÞ; ; gðx0; xqÞ satisfy gðx0; xjÞ4gðx0; xpÞ4 4gðx0; xqÞ, then we have the Grey relational order: xjxp xq.

When the grey relational coefﬁcient is conducted with respect to innovation policies, we then can derive the grade of the grey relation gðx0; xiÞ between the reference alternative gðx0; xiÞ ¼ 1 n Xn k¼1 gðx0ðkÞ; xiðkÞÞ. (10) 3.4. Cluster analysis

Cluster analysis, also known by the names of segmenta-tion analysis and taxonomy analysis, is a set of techniques for accomplishing the task of partitioning a set of objects into relatively homogeneous subsets, based upon inter-object similarities (Kachigan, 1991). The objective of cluster analysis is to group observations into clusters, such that each cluster is as homogeneous as possible with respect to the clustering variables. The first step in cluster analysis is to select a measure of similarity. Next, a decision is made on the type of clustering technique to be used. Third, the type of clustering method for the selected technique is selected. Fourth, a decision regarding the number of clusters is made. Finally, the cluster solution is interpreted (Sharma, 1996). To apply cluster analysis smoothly, we refined the cluster analysis procedure introduced byEveritt et al. (2001) and produced the essential definitions indicated below.

The basic data for most applications of cluster analysis is the usual n p multivariate data matrix, Y, containing the variable values describing each object to be clustered; that is Y ¼ y11 y12 y1p y₂₁ y₂₂ y_2p .. . .. . .. . .. . y_n1 y_n2 y_np 2 6 6 6 6 6 4 3 7 7 7 7 7 5 .

The entry yijin Y gives the value of the jth variable on

the ith object.

Initially, the data matrix, Y, is converted into an n n inter-object distance matrix D using the Euclidean distance

measure, where D ¼ d11 d12 d1n d21 d22 d2n .. . .. . .. . .. . dn1 dn2 dnn 2 6 6 6 6 4 3 7 7 7 7 5 and d2_ab¼X p j¼1 ðy_ajy_bjÞ2 (11)

where a and b stand for any two objects, and xaj, xbjstands

for the jth attribute of objects a and b, 1pjpp (Tzeng, 1978).

Once the inter-object distance matrix D is derived, the agglomerative procedure in the hierarchical clustering technique will be applied, by forming a series of partitions amongst the data: the ﬁrst consisting of n single-member ‘clusters’, and the last consisting of a single group containing all n individuals. At each stage, the methods fuse individuals or groups of individuals, which are closest (or most similar). Here, the distance between groups is deﬁned as that of the closest pair of individuals, where only pairs consisting of one individual from each group are considered, using the nearest-neighbor distance.

Deﬁnition 15. The distance between any two clusters, Cf

and Cg, can be calculated as

hfg¼minðdabja 2 Cf; b 2 CgÞ. (12)

4. SIP Mall industry background and current innovation policies

According to Worchel (2004), the market scale of SIP already exceeded one billion USD in 2004. Meanwhile, it is predicted that this market will continue steady growth and hit 1.8 billion USD by 2008. Despite the optimistic future of SIP, potential problems regarding SIP transactions that could hinder business have appeared. The complex problems include laws and regulations pertaining to SIP transactions, negotiations over SIP searching and usage rights, technical support including SIP integration into SOCs, SIP maintenance, and application engineering based upon current business models. Apparently, SIP transac-tions and integratransac-tions are not easy. e-Commerce business models for SIP transactions, which can effectively resolve the above-mentioned business, legal, and technical issues, will be most helpful to accelerate industry growth.

In the past, if an SIP user wanted to implement a SOC with SIPs integrated inside, the SIP user had to source the SIPs directly from different SIP providers and/or through different intermediates, including foundry, design service providers, brokers, and distributors, to acquire all the SIP solutions they needed (Fig. 4). Now users can purchase all the SIPs through a single web-based e-Commerce

(9)

mechanism, called an SIP Mall, and receive all SIP deliverables, business and legal services, and technical support (Fig. 5). When the SIP providers use the e-Commerce mechanism of the SIP Mall to source SIPs, negotiate business terms, define legal contracts, verify SIPs on-line in advance, and obtain pre-verified SIPs, the total time from design to market can be reduced significantly (Fig. 6). SIP providers also can reduce their time effectively through e-Commerce transactions. In the past, SIP provi-ders standardized, verified, and sold SIPs by themselves. After the establishment of the SIP Mall, SIP providers need only to upload their SIPs to the e-Commerce server, and the SIP Mall will finish the SIP simulations, verifications, marketing, sales, and customer services for them (Fig. 7).

The SIP Mall aims to develop a one-stop SIP shopping environment for worldwide customers, update SIP status in a timely manner, attract new business for wafer foundries, shorten time-to-tape-out of SIP customers, serve as an effective SIP marketing channel for worldwide SIP provi-ders, and make the e-Commerce mechanism easy to use.

To provide well-verified SIPs for next generation killer applications, the SIP Mall intends to provide SIPs to design houses, design service companies, and integrated device manufacturers (IDMs). At the same time, the e-Commerce mechanism will co-verify SIPs with foundries and define SIP technology roadmaps with system houses. As a powerful SIP transaction platform, the SIP Mall will serve as an efficient SIP marketing channel for worldwide SIP provi-ders. The SIP Mall will join industry standard definitions, so as to meet the SOC/SIP specification definition require-ments. For the software needed for the next generation of killer applications, the SIP Mall will cooperate with software providers to develop-related tools. To develop cost-efficient assembly and test solutions for next generation SOC products, the SIP Mall also will co-work with packaging and testing vendors. Finally, the SIP Mall will cooperate with professional educational institutes. On one hand, SIP Malls will provide SIP-related training. On the other hand, professional education institutions will provide professionals with this e-Commerce mechanism.

Fig. 4. SIP transactions in the past.

(10)

In addition to the United States, several other coun-tries—Japan, South Korea, France, Scotland, and Tai-wan—have rolled out SIP Mall programs. Following are brief introductions to each.

4.1. The United States

The Virtual Socket Interface Alliance (VSIA), at

www.vsi.org, is the United States’ SIP Mall program. It was formed in September of 1996 by a group of major EDA and semiconductor companies. They had two primary goals: ﬁrst, to establish a unifying vision for the chip industry; and second, to develop the technical standards required to enable the most critical component of that vision: the mix and match of SIP cores from multiple sources. VSIA later expanded that vision to meet the growing needs of the IC industry, by including software and hardware SIP for SOC design; and it recently has restructured the organization by creating a new type of working group, called a Pillar. Pillars address standards issues to improve productivity for SOC and IP developers and users. Currently, there are four Pillars. They cover SIP quality, SIP protection, SIP transfer, and R&D (VSIA, 2006).

4.2. Japan

The IP Trade Center (IPTC), at www.iptc.co.jp, was founded in 2000 as the only Japanese SIP Mall. It was

funded by several companies, including Toshiba, Canon, and Matsushita, among others. The mission of IPTC is creating and providing a communication and information-sharing mechanism to promote the reuse and exchange of SIP and design services, thereby enhancing the productivity of SOC and electronic equipment products. IPTC will realize collaboration of SIP and design service providers among member companies, in order to support the complex design and development processes of SOC (IPTC, 2004).

4.3. South Korea

System Integration & Intellectual Property Authoring Center (SIPAC), at www.sipac.org, was established in the KAIST in April 2001 to enhance the competitiveness of the Korean SOC industry, by establishing a web-based SIP trading system, a SIP database and an Internet-based SIP veriﬁcation and evaluation system, and by joining deﬁning SIP standards (Yoo, 2003).

4.4. France

The France SIP Mall, Design & Reuse (D&R), was founded in October 1997 as a web portal for value-added information in the ﬁeld of SIP and SOC information. D&R was funded by the Institute National Polytechnique de Grenoble (France) and the Medea project. D&R provides unique services to SIP providers by creating the best

Fig. 6. Concepts of SIP Malls saving customers’ time/cost to design and market.

(11)

communication channels for their customers. For this purpose, D&R holds private catalogs for the major SIP providers. Importing data from the SIP-provider database is automatic and data quality is guaranteed. These private catalogs also are used as an extension of the private micro sites being linked to the D&R portal. For SIP users, D&R rolled out a data intranet import gateway feature in 2001 for feeding users SIP and SOC resource catalogs. Based upon its early experience, D&R licenses technology and services to facilitate SIP sharing and reuse worldwide (Design and Reuse, 2006).

4.5. Scotland

Virtual Component Exchange (VCX) Software Ltd. was established and funded by the Scottish government in October 2000, as the non-proﬁt organization that launched the online VCX. The Exchange aimed at becoming the ﬁrst regulated buyer-to-buyer exchange for SIP. In September of 2003, VCX was spun off as an independent company to sell database solutions to both SIP buyers and sellers. The independent company also licenses SIPs through its server (Goering, 2003).

4.6. Taiwan

SIP Malls, including SIP Mall at www.sipmall.org and

Taiwan IP Mall at www.tipm.com, are operated by

Taiwanese design service companies and sponsored by the Taiwanese government’s Si-Soft project. Taiwan’s SIP Malls aim to provide an online SIP e-Commerce mechan-ism and SIP commercialization services, wherein users can purchase all SIPs through a single web portal and receive all SIP deliverables, business and legal services, and technical support. SIP providers also can reduce their time effectively through e-Commerce transactions and SIP commercialization services.

Comparing the innovation policies for developing SIP Malls in the above-mentioned six countries, the United States and Japanese SIP Mall programs were sponsored by major semiconductor companies. The United States and Japanese governments did not take significant roles in developing their SIP Malls. Regarding the other four SIP Malls, in South Korea, France, Scotland, and Taiwan, the major applications of environmental innovation policy tools for developing their programs were almost the same. The governments were involved by funding the SIP Mall program either directly (e.g., Scotland), through govern-ment sponsored R&D programs (e.g., Taiwan), or through governmental R&D institutes (e.g., South Korea and France). For the supply side policy, South Korea and Taiwan leveraged almost the same tools, including scientific and technical, education as well as information, of which platform and SIP development, plus SIP qualification standard definition by governmental R&D institutes, was the most significant. On the other hand, supply side policy tools of France and Scotland were

insigniﬁcant. Finally, no country leveraged demand side policy tools as the major approach for developing their SIP Mall programs.

5. An empirical study: the case for Taiwan’s SIP Mall industry

The current innovation policy being adopted by the Taiwanese government includes: (1) alleviation of taxation; (2) loan subsidies; (3) supply of information and techno-logical assistance institutions; (4) government procure-ment; (5) protection of research results; and (6) cultivation of manpower (Shyu and Chiu, 2002). Meanwhile, to support the development of Taiwan’s SIP and SIP Mall industry, the government has leveraged several policy tools speciﬁcally targeting the SOC and SIP industry.

Even though the government has exerted considerable effort into developing the SIP Mall industry over the past 5 years, Taiwanese SIP Malls still are in the exploratory phase, while SIP sources are limited and SIP transactions through SIP Malls remain few. Meanwhile, Taiwanese SIP Malls are being operated by design service companies. When SIP Mall operations compete for engineering resources needed for customer projects, design service companies always set the priorities of SIP Mall operations as lower priorities, and this has caused SIP Mall inefﬁciency.

How to deﬁne innovation policies that can enhance the innovation capability and competitiveness of Taiwan’s SIP Mall industry and enrich Taiwan’s star SIP portfolio has become one of the most important issues amongst policy makers.

The authors propose an analytical framework for defining the innovation policy portfolio of Taiwan’s SIP Mall industry, by first analyzing current IIRs by means of summarizing opinions from experts using the Delphi method, and then simplifying the complex IIRs using DEMATEL. Finally, the relationships between the simpli-fied IIRs and the innovation policy tools were derived via GRA. Six experts within Taiwan’s SIP Mall industry were invited to provide their inputs at each stage of this empirical analysis. Both the details and results of the procedures are presented below.

5.1. IIRs derived by Delphi

Twenty-four IIRs needed for industrial innovation of Taiwan’s SIP Mall industry were collected by interviewing experts in government and Taiwan’s SIP Mall industry using Delphi procedures. Descriptions of these IIRs as follows:

(1) e-Commerce capability: SIP e-Commerce capabilities include the establishment of an e-Commerce infra-structure, public policies and regulations, electronic data interchanges, on-line payments, and security mechanisms. These capabilities deﬁnitely will be

(12)

critical to deﬁne and enable e-Commerce mechanisms, shorten the time to market for SIPs, and enable SIP business.

(2) Market leadership and customer education: Technology push is becoming the marketing trend in the IC and SOC industries. Those who have the capacity to educate their customers will win at the end. Market leaders then can define specifications for their next-generation products. With higher market shares and profits from already developed products, market leaders also can generate returns from investments sooner than their competitors. Consequently, more and more resources can be invested into innovation activities to accelerate the time-to-market for next generation products. Thus, market leaders can con-firm their situation even longer and to a greater degree.

(3) Time-to-market capability: Time-to-market always is the most important factor for vendors seeking to help customers achieve time-to-money. As the complexity of SOCs increases, the ability of vendors to roll out solutions in the shortest time possible is crucial for maintaining customer relationships.

(4) SIP sourcing capability: In the SOC era, SIP sourcing has become one of the most important issues in SOC design, since SIP design and reuse already have become one of the most frequent activities. Effective sourcing of SIP can shorten the time-to-design. (5) SIP integration service capability: SIP integration

services assist customers implement designs from concept-to-production within the shortest time. SIP integration services also help customers to narrow the gap between system companies and silicon foundries, to enable process technology deployment, and to shorten time-to-design, and therefore, time-to-reven-ue. For system houses and medium-to small-sized design houses, SIP integration services help customers to ﬁnish their SOCs.

(6) SIP qualification capability: Well-qualified SIP cores are necessary for SOC success, both in terms of engineering and business. Thus, SIP qualification capabilities of SIP providers are essential for innova-tion. Typical SIP qualification includes inspections of SIP documentations, specifications, SIP verification results, and SIP pilot run results.

(7) SIP verification capability: SIP verification is the design procedure used to verify the functions and timing of SIP cores. Thus, SIP verification capability is as important as SIP qualification with respect to guaranteeing the final success of an SIP.

(8) Multiple Project Wafer (MPW) services: As technol-ogy evolves, mask and engineering run costs grow exponentially with each process generation. MPW services aim to reduce IC and SIP developer costs, so as to develop and facilitate faster prototyping by sharing the costs of a common mask set and engineering-run. Moreover, these services encourages

innovation, and offers their customers a great opportunity to prove their design, product and market test samples.

(9) Legal protections: The SIP business is built on the foundation of IP. Naturally, security is one of the main concerns in the SIP business. One way to secure IP is by using legal procedures, as previously mentioned. However, such methods are not iron clad. As IP trading moves further onto the Internet, better security methods will be needed (Gajski et al., 2000). (10) Funding capability: IC and SIP designs have become more and more complicated in advanced processes. The most up-to-date IC design flows and EDA tools are required to achieve higher speed, lower power, and lower cost. MPW blocks in advanced processes are more expensive. Meanwhile, setting up the e-Com-merce infrastructure requires greater investments in servers and firewalls. How to leverage various funding tools, including loans, venture capital investments, initial public offerings (IPOs), and changeable bonds (CBs) has become the most important factor influen-cing operation efficiency.

(11) Establishment of SIP know-how: Domain knowledge regarding SIP design speciﬁcation, integration, testing, mass production, and application can guarantee proper SIP applications and transactions.

(12) Establishment of electronic systems know-how: SOC needs a great deal of electronic system know-how to achieve the ﬁnal successful integration of SIP applica-tions into chips and electronic systems. The establish-ment of electronic system know-how also is important for SIP success in engineering and business.

(13) Joint development capability: Individual organizations no longer can rely on their own resources to compete in today’s world. Rather, they should look for strategic interactions that will allow them to leverage internal resources effectively, by investing in core competencies and contracting out other knowledge domains (Sobrero and Roberts, 2002). Various speciﬁcations for embedded processors, peripherals, memories, analog, and mixed signal cores are emer-ging as future needs for 3C (computer, communica-tion, and computer) applications. There is almost no single company—no matter how professional an SIP provider, fabless IC design house, semiconductor foundry, or even an IDM is—that can afford the human resources and R&D capabilities necessary to develop every SIP, especially at a time when the time-to-market of SOC products is shortening rapidly. Hence, a joint development capability (including technology transfer) has become a key aspect of SIP innovation.

(14) SIP porting capability: Moving an SIP to the existing process of another semiconductor fab always involves redesign and extensive product development, and often multiple expensive iterations. A better way is to ‘port’ the process; i.e., move it cell by cell without

(13)

changing the SIP created to use it. By adopting the basic methods used in the silicon industry and paying attention to process details that affect performance, successful porting is faster, much less expensive, and less disruptive to the flow of the product. Porting also results in multiple qualified fab sites for a given SIP. An added benefit is that multiple geographically separated production facilities make the supply chain immune to disruption by natural disaster or power interruptions. Porting also eliminates the need for customers to re-qualify a product containing chips from a new source (Williams et al., 2002).

(15) Strategic alliance capability: Strategic alliances typi-cally take the form of an agreement between two or more ﬁrms to work together for a speciﬁc project, and tend to focus on near-market development projects (Tidd and Brocklehurst, 1993). SIP Malls should have strong strategic alliances with research institutes, universities, design service companies, and system houses to access the most up-to-date techniques and domain know-how regarding system application en-gineering.

(16) Close relationship with foundries: SIPs, especially mixed signal and analog SIPs from professional SIP providers, always target specific processes of major foundries—for example, TSMC, UMC or IBM—to target major foundry SOC customers in advanced processes. Close relationships with foundries are essential for SIP Malls, so that they can have a SIP verification, qualification and marketing channel. (17) Close relationships with design service companies: Design

service companies provide SOC integration, MPW, and SOC turnkey services, which may serve as bridges between customers and SIP Malls. A close relationship with design service companies may guarantee SIP integration, veriﬁcation, and qualiﬁcation success. (18) Close relationships with customers: Customer

relation-ship management is a strategy used to learn more about customers’ needs and behaviors, so as to develop stronger relationships with them.

(19) Close relationships with SIP providers: The close relationship between an SIP Mall and SIP providers may guarantee that the SIP Mall accesses the most up-to-date SIP information, including SIP speciﬁca-tions, simulation models, product roadmaps, and business models. This information may assist custo-mers in implementing their SOCs correctly, using the most advanced SIPs and design information.

(20) Marketing capability: Strong SIP Mall marketing capability includes SIP product and service deﬁni-tions, pricing, promotion and advertisement, compe-titive analysis, and more.

(21) Design service capabilities: The design service capabil-ities of SOC refers to the capabilcapabil-ities to provide customers who ﬁnish SOC speciﬁcation design or SOC circuit design with the remaining procedures or processes required for SOC commercialization. These

detailed procedures include front-end design, back-end (place and layout), and turnkey (tape-out of the layout to semiconductor wafer fab for SOC fabrica-tion) services. The above-mentioned design service capabilities deﬁnitely are helpful for SIP Mall success, since strong design service capability implies strong SIP integration capability, a key factor in SOC success.

(22) SIP safeguard techniques: SIP safeguard techniques mean watermarks and other approaches implemented in SIPs to prevent the illegal use or duplication of SIPs. (23) R&D knowledge management capability: The modern

R&D system magniﬁes the role of information technology (Barthe’s and Tacla, 2002; Dennis et al., 1998) and emphasizes the platform and architecture of the whole R&D system. In this regard, knowledge management (KM) becomes an indispensable requi-site of the fourth generation R&D. In fact, the R&D process can be considered primarily a knowledge management process, because it transforms informa-tion on technological advancements and market demands into knowledge. That knowledge then can be used to develop new product concepts and process designs (Kerssens-Van et al., 1996;Nieto, 2002; Park and Kim, 2004).

(24) Human resources management capability: ICs have become one of the fastest growing and most highly invested industries in Taiwan. In order to keep pace with the technological advancements in nanometer semiconductor manufacturing, IC design and SOC techniques also are evolving. This evolution is leading to a strong demand from the industry for a large number of well-trained IC design engineers, as well as supportive human resources in business and law (Advisor Ofﬁce of Ministry of Education, 2004).

5.2. IIRs reduction by DEMATEL

Since the 24 IIRs summarized using the above Delphi process seem too complicated to be analyzed, the key innovation requirements will be deducted using the DEMATEL method that was introduced in Section 3.2. First, the direct relation/influence matrix A is introduced, as shown inFig. 8. After that, the direct relation/influence matrix A is normalized, based upon Eq. (1) and the normalized direct relation/influence matrix N is shown in

Fig. 9. Finally, the direct/indirect matrix is deducted, based on Eq. (3) and shown inFig. 10. The causal diagram of the total relationships is shown inFig. 11where major IIRs are deducted by setting the threshold value at 0.10. The symbols being used to represent the IIRs are summarized inTable 2. It is obvious that the major IIRs are IIR8, IIR9, IIR11, IIR12, IIR15, IIR16, IIR20, IIR22, IIR23, and IIR24, based upon the values of (rici). These values stand

for MPW Services (IIR8), legal protections (IIR9), SIP know-how establishment (IIR11), electronic systems

(14)

know-how establishment (IIR12), strategic alliance cap-ability (IIR15), close relationship with foundries (IIR16), marketing capability (IIR20), SIP safeguard techniques (IIR22), knowledge management capability (IIR23), and human resources management capability (IIR24).

5.3. Mapping-derived IIRs to innovation policy tools using GRA

GRA is used here to derive the relationships between IIRs and innovation policy tools. The initial relationship

matrix G, a 12 10 matrix that maps the relationship between 12 innovation policy tools and 10 major IIRs, was obtained by collecting opinions from experts (Fig. 12). The symbols being used to represent the policy tools are introduced inTable 3. The normalized relationship matrix X then was derived based upon Eqs. 7 and 8 (Fig. 13).

Let the value of each element of x0, the reference pattern,

be ‘1,’ which stands for the best case for the elements in the normalized correlation matrix. By setting the distinguished coefﬁcient z as 0.5, the Grey relation coefﬁcients were derived using Eq. (9). Finally, the grades of Grey relation

Fig. 8. The direct relation/inﬂuence matrix A.

(15)

gðx0; xiÞ were derived using Eq. (10) (see Table 4). We established the result that

P2 P3 P1 P4 P7 P5 P6 P8 P10 P9 P12 P11,

which stands for the policy priorities for developing the SIP industry.

To cluster the above policy tools so as to establish the policy portfolios, a 1 12 multivariate data matrix Y containing the Grey grades was established as

Y ¼ 0:514½ 0:833 0:784 0:500 0:489 0:482 0:500 0:411 0:397 0:394 0:400 0:399.

The data matrix Y was split into three clusters using SPSS. The three clusters—{P2, P3}, {P1, P4, P7, P5, P6} and {P8, P10, P9, P12, P11}—stand for the most important, important, and not important policy portfolios, respectively, in terms of developing Taiwan’s SIP Mall industry (Fig. 14). Here, scientific and technical (P2) and education (P3) are the most important policy tools, while public enterprise (P1), information (P4), legal and regula-tory (P7), financial (P5) and taxation (P6) also are important policy portfolios needed to develop Taiwan’s SIP Mall industry. Other policy tools—including political (P8), public services (P10), commercial (P11), and overseas agent (P12), and procurement (P9)—are classified as less important policy tools.

Fig. 10. The direct/indirect matrix T.

-1.000 -0.500 0.000 0.500 1.000 1.500 2.000 2.000 2.500 3.000 3.500 4.000 4.500 D+R D-R

(16)

6. Discussion

Despite conceptual advances, econometric models, and illustrative case studies, there are no straightforward

answers to the questions: what elements should an innovation policy include, and how should such policies be implemented (Salmenkaita and Salo, 2002)? Examples of deﬁning innovation policy elements with traditional qualitative approaches include the national innovation policy proposal for Cyprus byHadjimanolis and Dickson (2001), the innovation policy proposal for developing

Taiwan’s competitive advantage by Shyu and Chiu

(2002), and the reconﬁguration of national innovation systems for German biotechnology by Kaiser and Prange (2004). Although popular, traditional qualitative ap-proaches could be subjective and misleading. Meanwhile, collected information based on the assumption of inde-pendence between the various innovation requirements could be time wasting during the analysis procedure. Moreover, the vague correlations between innovation requirements and policy tools, and the lack of priorities of the policy tools deducted from the analysis procedure could cause governments to waste resources, merely because either low or no-priority policy instruments are executed.

In this paper, we demonstrated that an analytical approach can identify the complex relationships that exist between the various IIRs, and can clearly deﬁne the correlations between IIRs and innovation policy tools, thereby ranking policy priorities. DEMATEL served as an efﬁcient tool to deduct the cause and effect relationships within an MCDM problem structure, and to simplify the relationships between IIRs. Even though the IIRs being

Table 2

Symbols being used for representing the IIRs

Symbol Contents

IIR1 e-Commerce capability

IIR2 Market leadership and customer education IIR3 Time to market capability

IIR4 SIP sourcing capability IIR5 SIP integration service capability IIR6 SIP qualiﬁcation capability IIR7 SIP veriﬁcation capability

IIR8 MPW services

IIR9 Legal protections IIR10 Funding capability

IIR11 SIP know-how establishment

IIR12 Electronic systems know-how establishment IIR13 Joint development capability

IIR14 SIP porting capability IIR15 Strategic alliance capability IIR16 Close relationship with foundries

IIR17 Close relationship with design service companies IIR18 Close relationship with customers

IIR19 Close relationship with SIP providers IIR20 Marketing capability

IIR21 Design service capabilities IIR22 SIP safeguard techniques

IIR23 Knowledge management capability IIR24 Human resources management capability

Table 3

Symbols being used for representing the policy tools Symbol Contents Symbol Contents

P1 Public enterprise P7 Legal and regulatory P2 Scientiﬁc and technical P8 Political

P3 Education P9 Procurement

P4 Information P10 Public services

P5 Financial P11 Commercial

P6 Taxation P12 Overseas agent

Fig. 13. The normalized correlation matrix X.

Fig. 12. The initial relationship matrix G.

Table 4

The grades of Grey relation with respect to innovation policy tools (z ¼ 0.5)

Innovation policy tool Grade Innovation policy tool Grade Public enterprise 0.514 Legal and regulatory 0.500 Scientiﬁc and technical 0.833 Political 0.411

Education 0.784 Procurement 0.397

Information 0.500 Public services 0.394

Financial 0.489 Commercial 0.400

(17)

deducted by DEMATEL in Section 5.2—e-Commerce capability (IIR1); market leadership and customer educa-tion (IIR2); time-to-market capability (IIR3); SIP sourcing capability (IIR4); funding capability (IIR10); joint deve-lopment capability (IIR13); SIP porting capability (IIR14); strategic alliance capability (IIR15); close relationships with customers (IIR18); close relationships with SIP providers (IIR19); and design service capabilities (IIR21)—were recognized by the experts as essential to enabling industrial innovation, the IIRs were classiﬁed as to the effects of major IIRs. The threshold value was set at 0.100, which covers 69 relationships, which comprise approximately 12% of the total 576 (24 24) relationships. If the threshold value was set higher than 0.100, the number of relationships to be analyzed could be too few to be representative; some IIRs would not be included in the total relationship. On the other hand, if the threshold value was set at a lower threshold value, the relationship number would be too huge to be analyzed. Thus, based upon a threshold value of 0.100, the relationships between IIRs— classiﬁed as major IIRs in the cause group and other IIRs in the effect group, as well as those either deducted or neglected in the GRA procedures—are listed below in

Table 5. Table 5 shows both the clear cause-and-effect relationships and the rationales behind these relation-ships, demonstrating that every IIR deducted by DEMA-TEL already has been impacted by one or more major IIR(s).

For multiple criteria decision-making (MCDM) pro-blems, GRA has advantages over traditional statistical correlation analysis. GRA is not constrained by sample numbers, whereas the total number of samples should be over 30 for traditional statistical relational analysis. Meanwhile, the weights for criteria can be introduced while calculating the grade in GRA (however, to facilitate the survey, we assumed that the weights of each criterion are equal.) Moreover, GRA derives the grade of Grey relation between criteria (IIR) and alternatives (innovation policy instruments), whereas statistical relational analysis derives the pair-wise correlation coefﬁcients only. Finally, with GRA, the concepts of minimum and maximum are introduced, and the reference point is set for calculating the grades, speciﬁcally the distances between the variable and the reference point, while statistical correlation analysis calculates the sum of the products of two standardized values ðzi¼ ðxij¯xiÞ=si; i ¼ 1; 2; . . . ; pÞ only.

Fig. 14. Classiﬁcation of the grades with respect to each policy tools by cluster analysis.

Table 5

The cause and effect relationships derived by DEMATEL

Effect Causes Cause and effect relationships and rationale

IIR1 IIR24 Human resources management (IIR24) affects the SIP Malls e-commerce capability IIR2 IIR1, 6, 7, 8, 10, 11, 14,

15, 19, 20, 24

e-Commerce capability, SIP technical capabilities (IIR 6-7), funding capability (IIR10), strategic alliance capabilities (IIR15), relationships with providers (IIR19), human resource management (IIR24), and etc. affect market leadership and customer education

IIR3 IIR4, 5, 7, 11, 19, 24 SIP technical capabilities (IIR 4, 5, and 7), SIP know-how establishment (IIR11), relationships with providers (IIR19), and human resource management (IIR24) affect time to market capability IIR4 IIR24 Human resources management (IIR24) affects SIP sourcing capability

IIR5 IIR11, 24 SIP know-how establishment (IIR11) and human resource management (IIR24) affects SIP integration service capability

IIR6 IIR24 Human resources management (IIR24) affects SIP qualiﬁcation capability IIR7 IIR24 Human resources management (IIR24) affects SIP veriﬁcation capability

IIR10 IIR 7, 24 SIP veriﬁcation capabilities (IIR7) and human resources management (IIR24) affect funding capability

IIR13 IIR24 Human resources management (IIR24) affects joint development capability IIR14 IIR24 Human resource management (IIR24) affects SIP porting capability

IIR17 IIR24 Human resource management (IIR24) affects close relationship with design service companies IIR18 IIR24 Human resources (IIR24) is the major factor affecting the relationship with customers (IIR18) IIR19 IIR 11, 24 SIP know-how establishment and human resources management are major factors affecting the

relationship with SIP providers

IIR21 IIR24 Human resources (IIR24) is the major factor affecting SIP Mall’s design service capability Remark: Threshold value ¼ 0.100.

(18)

Apparently, GRA serves as an effective tool for analyzing the correlation between policy tools or strategies versus IIRs or any criteria that are necessary for evaluating policies or strategies, especially when available experts in some specific fields of emerging technologies are limited. Thus, this procedure provides a solution for addressing the need to define a policy portfolio with clear steps, and is more appropriate for real-world policy definition problems.

Even though the Taiwanese government executed several national projects and policies to support the development of SIP Malls in the national Si-Soft project, which was rolled out in 2002, human resources management cap-ability, knowledge management capcap-ability, SIP porting capability, MPW services, SIP know-how establishment, marketing capability, SIP safeguard techniques, strategic alliance capability, SIP veriﬁcation capability, electronic systems know-how establishment, close relationship with design service companies, legal protections, e-Commerce capability, close relationship with foundries, funding capability, and SIP sourcing capability remain factors needed for industrial innovation.

Thus, innovation policy recommendations for develop-ing Taiwan’s SIP Mall industry, derived from analyzdevelop-ing the Grey relationships between IIRs and innovation policy tools, are discussed below.

Science and technical (P2) and education (P3) policy tools are classified as the most important tools. Science and technical policy tools (P2)—including research institutes like the Industrial Technology Research Institute (ITRI) and the National Chip Implementation Center (CIC) of Taiwan should serve as providers of expensive advanced-technology MPWs for SIP Mall and SIP vendors (IIR8). The research institutes also should enhance training in domain know-how of new SIP techniques (IIR11) and advanced electronic systems (IIR12). These tools can serve as ‘bridge institutes’ between SIP Malls and foundries and SIP providers (IIR15, 16) for strategic alliances. ITRI also should define the national SIP roadmap, and develop SIP safeguard techniques (IIR22) and a SIP Mall knowledge management platform (IIR23) as references or transferable techniques for SIP Malls. Universities (P3) should train more personnel and enhance courses in the above-noted fields, including MPW operation, SIP and electronic system know-how, safeguard techniques, high-technology marketing, and knowledge management, in order to generate the personnel to meet the future needs of SIP Mall operations.

Public enterprise (P1), chosen as an important policy tool, implies that the current status—wherein existing SIP Malls still are experimental projects, operated either by design service companies or by non-profit research institutes that are not efficient and cannot generate profit—is not acceptable. To spin off existing SIP Malls as individual companies at a time when SIP Malls still are not profitable, the government should participate in SIP Mall companies via direct investment. With such govern-mental backing, SIP Mall companies would be more capable of establishing relationships with leading foundries

and forming strategic alliances with major SIP providers and IC industry leaders, as SIP Mall customers.

Meanwhile, the government should leverage its financial (P5) tools through continuous funding of the SIP Malls, and by providing low interest loans. Doing so will meet the requirements that SIP Malls purchase expensive advance processes, MPW blocks, or wafers for MPW services (IIR8) to aid in SIP verification and qualifications. It also would allow for SIP Mall employee salaries (IIR24) and EDA tool investments to be sufficient. Taxation holiday or tax incentives (P6) also are regarded to be important tools for MPW investments (IIR8) and sustaining high-technol-ogy human resources (IIR24).

Currently, SIP Malls usually deﬁne their SIP portfolios and roadmaps based upon the domain knowledge of the design service companies operating them. Research in-stitutes (P2) like the SOC Technology Center (STC) of ITRI and information centers and advisory and consul-tancy services (P4), like the Taiwanese government-funded Market Intelligence Center (MIC) of the Institute for Information Industry and the Industrial Economics & Knowledge Center (IEK) of ITRI, can provide technology and market information needed for an SIP development roadmap and SIP pricing, thereby forming the basis of SIP marketing jobs (IIR20).

We also would suggest that the Taiwanese government protects SIP transactions online and continues its efforts in SIP protection through legislation (P7), especially in the related ﬁelds of SIP source codes, layout, and software transmission using the Internet, so as to enhance the legal protection of SIP transactions (IIR9) on the web.

Comparing the proposed innovation policy portfolio with the current innovation policy, it is apparent that the Taiwanese government should participate in SIP Mall companies via direct investment (P1) rather than through funding R&D projects only. Major scientiﬁc and technical institutes (P2), including ITRI and CIC, proactively should provide SIP management courses, human resources, and design and engineering services; they also should serve as one of the major technology sources, and as bridging institutes between SIP providers, system designers, fabless design houses, foundries, and so on. Meanwhile, the human resource supply and demand gap remains signiﬁ-cant, even though the Si-Soft Project tried to close the human resource supply and demand gap by using universities as a resource (P3). STC, IEK, and MIC should proactively provide technical, alliance, and free marketing information (P4) to new SIP Malls, which cannot afford expensive information services. Finally, the government’s continuous funding (P5), tax allowances (P6), protecting SIP transactions online and continuous efforts in SIP protection (P7) also are necessary (seeTable 6).

7. Conclusions

Despite the fact that SIP Malls play an important role in future SIP industry development, existing SIP Malls have not