資料包絡分析法應用於光儲存產業供應鏈中聯盟對象選擇之研究

行政院國家科學委員會專題研究計畫 成果報告

資料包絡分析法應用於光儲存產業供應鏈中聯盟對象選擇

之研究

研究成果報告(精簡版)

計 畫 類 別 ： 個別型

計 畫 編 號 ： NSC 97-2221-E-151-034-

執 行 期 間 ： 97 年 08 月 01 日至 98 年 09 月 15 日

執 行 單 位 ： 國立高雄應用科技大學工業工程管理系

計 畫 主 持 人 ： 王嘉男

計畫參與人員： 碩士班研究生-兼任助理人員：陳立勤

碩士班研究生-兼任助理人員：黃其恆

報 告 附 件 ： 出席國際會議研究心得報告及發表論文

處 理 方 式 ： 本計畫涉及專利或其他智慧財產權，2 年後可公開查詢

中 華 民 國 98 年 10 月 14 日

行政院國家科學委員會補助專題研究計畫成果報告

資料包絡分析法應用於光儲存產業供應鏈中聯

盟對象選擇之研究

計畫類別：■個別型計畫

□整合型計畫

計畫編號：NSC97－2221－E－154－034－

執行期間：97 年 8 月 1 日 至 98 年 7 月 31 日

計畫主持人：

王嘉男

共同主持人：

計畫參與人員：

陳立勤、黃其恆

執行單位：

國立高雄應用科技大學

中

華

民

國

9 8

年

9

月

3 0

日

行政院國家科學委員會專題研究計畫成果報告

資料包絡分析法應用於光儲存產業供應鏈中聯盟對象選擇之

研究

A DEA-base method for candidate section of supply chain

management in compact disc industry

計 畫 編 號：NSC 97 － 2221 － E － 154 － 034

執 行 期 限：97 年 8 月 1 日至 98 年 7 月 31 日

主

持

人：王嘉男

國立高雄應用科技大學，工業工程與管系

共 同 主 持 人：

計畫參與人員：陳立勤、黃其恆

國立高雄應用科技大學，工業工程與管系

一、中文摘要

全球光儲存生產廠需求量增加競爭劇烈的環

境下，如何在供應鏈(Supply Chain)中，以策

略聯盟

(Strategy Management)創造出競爭優

勢，已成重要課題。本研究以資料包絡分析法

(Data Envelopment Analysis，簡稱DEA)為基

礎，提出一套有系統的模式，其目的是在為光

儲存產業，在供應鏈中進行策略聯盟時，能有

效尋找最佳策略聯盟的夥伴，並分析聯盟合作

後新公司的預期經營效率與改善的方案，以增

進企業的競爭優勢。本研究以台灣光儲存產業

實際的資料來研究，本研究期望提出一套有效

的模式，能應用於光儲存產業中，各企業間供

應鏈策略聯盟合作之評估，也能為管理者評估

選擇聯盟對象時提供有效的分析與建議。建

議。

關鍵詞：資料包絡分析法、

DEA、策略聯盟、

光碟片

Abstract

In recent competitive business environment,

increasing

competition

becomes

a

very

significant issue for compact disc industry.

Many companies are looking forward to strategy

alliance in supply chain management (SCM) for

enhancing their capability and realizing their

business

roadmap.

According

to

data

envelopment analysis (DEA) and heuristic

method, the objective of this paper is to develop

an effective method to assist enterprise to

evaluate the operation efficiency and find the

candidate priority of strategy alliance under

several different inputs and outputs for the

partner of supply chain. Moreover, analyze

resources reallocation after alliance. This could

be great contribution for industrial managers

when choose the partner of supply chain.

Keywords: DEA, Alliance, Heuristic Algorithm

二、緣由與目的

越高，所以對儲存容量的需求也越來越高，而

面對技術不斷的研發，未來對容量的需求將不

斷的升高，目前有兩大陣營想要取代現有的

DVD 技術，兩者推動的標準容量和目標群都

不同，一者是藍光碟片協會，

Blu-ray（藍光

DVD）是 50GB，著重 DVD 片子可以做為大

量資料的備份儲存媒體，如戴爾（Dell）、惠

普（HP）、日立（Hitachi）、LG、Matsushita、

Mitsubishi、Pioneer、皇家飛利浦（Philips）、

三星（Samsung）、Sharp、Sony、TDK，以及

Thomson 多媒體等。其中 Sony、Sharp、三星

等廠商，已經推出支援藍光碟片的播放機；

Sony、三星、Panasonic 則在近期展出藍光碟

片。一者則是

DVD 論壇（DVD Forum）

，

DVD

論壇發展的是

HD DVD（紅光光碟）容量為

30GB，強調大容量可以存取高畫質、解析度

高的影片，雖然沒有藍光那麼好的儲存容量，

但因為成本較低，東芝（Toshiba）、NEC 、

三洋（Sanyo）則是 HD DVD 標準的最大支持

者。

面對如此激烈的競爭，要如何有效的改進

並創造出有利的競爭優勢，評估其策略聯盟對

象之選擇與預估聯盟後的經營績效，是光碟片

進入下一個世代前所勢在必行的，而在未來光

碟片產業中的廠商，要如何的發展以及運用何

種方式來增加自己在產業中的競爭力，並提升

產業整體的經營效率是本研究的目標，因此本

文運用

DEA 提出的評估模式在以策略與效率

角度訂定光碟片產業聯盟之策略來探討未來

光碟片產業合作的模式，首先以

DEA 模式分

析所有候選公司的效率，且在執行產業間聯盟

時，使用

DEA 計算各企業策略聯盟前後的個

體經營效率，以歸納出企業間聯盟的最佳組

合，和分析聯盟後新公司的預期經營效率，最

後提出資源分配及整合的改善建議。

本研究的研究架構簡述分成五個章節來

進行研究相關之討論， 第壹章：敘述本文之

研究動機、研究目的與研究架構。第貳章：介

紹資料包絡分析法與探討本文相關之文獻。第

參章：資料的來源、研究方法的的說明介紹。

第肆章：運用本文所提出評估模式作實際演

繹，並將數據結果作適當之分析研究。第伍

章：將本研究之貢獻做完整說明、並舉出未來

研究方向及建議。

三、研究報告應含的內容

資 料 包 絡 分 析 法

(Data Envelopment

Analysis)也就是所謂的DEA 分析方法，是由

Charnes, Cooper, Rhodes (1978)三位學者所提

出 的 概 念 ， 三 位 學 者 主 要 是 受 到 了

Farrell

(1957)所提出的生產效率之衡量所影響。而資

料包絡法是一種無母數效率前緣的分析方

法，此方法主要是用衡量多項投入與多項產出

之決策制定單位（Decision Making Unit,DMU）

的一種相對效率方法。也就是所謂的效率評估

之方法。主要使用了數學規劃的技巧求出效率

前緣（Efficiency frontier）

，再應用此效率前緣

來衡量各決策制定單位的生產效率。它是一種

由觀測值以前緣方法加以包絡，在經濟學上所

指之所有可能最佳之解析之點所組成，且形成

一條包絡線，稱為效率前緣。而落在效率前緣

上的決策單位被認為其投入及產出是有效率

的，而落在效率前緣下方的則被認為是不具效

率的。

DEA 模 式 最 為 常 人所使 用 的 為 CCR 、

BCC、CCSS、CCGSS 四種模式。CCR模式

為Charnes, Cooper and Rhodes （1978）所提

出 的 ；

BCC 模 式 則 為 Banker, Charnes and

Cooper（1984）所發展出的；CCSS 模式是由

Charnes（1983）等四位學者提出；CCGSS 模

型是由Charnes（1985） 等人於1985 年提出

的。本研究則只針對CCR 模式與BCC 模式做

一說明。

Farrell及Fieldhouse（1962）：擴充Farrell

（1957）固定規模報酬之假設至規模報酬增模

式 ， 建 立 效 率 衡 量 基 本 理 論 。

Charnes,

Copper,Lewin, Morey及Roussean（1985）：首

先對DEA之敏感度提出分析。Jemric et al.：新

銀行的效率較舊銀行的效率來的高；小型銀行

通常較具有全球化的效率，大型銀行在規模報

酬變動下的效率卻是較高的。

Bhattacharyya et

al. (1997)：公營銀行比民營銀行較有經營效

率；外商銀行在初期經營效率顯著增加，且外

商銀行能適應競爭增加之環境，反而公營銀行

適應較差。

目前已發表的國內外之論文，多半是以

「績效評估」為主題；而以DEA為評估分析「策

略聯盟」的主題文章並不普遍，因此本研究希

望以建構符合企業需求的評估工具，提供管理

者在選擇評估策略聯盟對象時，企業間聯盟的

最佳組合，和分析聯盟後新公司的預期經營效

率，最後提出資源分配及整合的改善建議。

壹、研究方法與流程

研究目前DEA的相關文獻，並以相關產業

之資料首先作一個收集，之後在運用DEA分析

法來對候補廠商以及個案公司來作一個績效

評估，以一個策略聯盟的模式，來達成本研究

之目的。

流程說明：

步驟

1 資料收集

1-1 相關資料收集：

本研究以光碟片產業作為主題，並蒐集目

前台灣光碟片產業中上市上櫃的全部廠商共

9 家來作聯盟的分析。

1-2 建立 DEA 效率評估模式:

在本論文中所選擇的模式為

CCR 模式，

使用

CCR 模式進行效率評估，因為本研究的

目的為判斷企業在策略聯盟時所該選擇的對

象，而在評估的過程中，我們將其分級，不需

要針對企業的規模效率作出建議增減投入產

出項的建議，所以我們選擇使用

CCR 模式，

來求得企業的總效率僅可。

步驟

2 選擇評估公司

界定產業未來在策略聯盟時，如何選擇正

確的合作伙伴的規則。本研究在分析過後，決

定以個案公司

G 公司來研究。

步驟

3 使用相關係數分析決定投入產出變數

不論是什麼樣的產業，在未來的長期發展

下，都應該是全方面的考慮所有的構面，而不

應該侷限於獲利上，這樣對企業來說，只有短

暫的發展，而無長期的計畫，因此，本研究考

慮到光碟片產業的特性，並利用相關係數分析

法來作分析六個投入產出，最後根據相關係數

分析的結果，決定使用投入指標的資本額與研

發經費跟產出指標的營業收入及營業淨利。

步驟

4 使用 DEA 分析所有候選公司

以CCR-I模式將候選公司做效率評估排

名，正常來說，績效高的廠商對於合作聯盟的

意願通常都不高，而排名較後面的產業合作聯

盟的意願則會比較高，在下一個步驟裡將加入

個案公司，再作一次效率評估。

步驟

5 加入個案公司

再次以CCR-I模式將「個案公司+候選公

司」(假設為策略聯盟合作模式) 將其模擬為

一個新企業，再與所有候選公司做效率評估排

名。

步驟

6 以 DEA 作效率分層排名

首 先 在步 驟四 對 候 選公 司 來作 績效 評

估，並加以排序，在步驟五加入個案公司後與

其他企業再作一次績效評估。

步驟

7 整體效率分析+差額變數分析

將以上數據資料運用整體效率來進行分

析，針對排名提升者優先推薦為策略聯盟之伙

伴，而效率前緣未達

1 者或效率排名持平或降

低者，利用差額變數分析來規劃改善方案，並

對策略聯盟後的公司進行分析及討論。

綜合上述步驟，本研究以策略聯盟夥伴之

現有最新資料，選取2003年的年度報表資料，

而在資料廣度上，一些主要指標在取得不易的

狀況下，並未能納入分析資料中，如資產總

額、營業毛利等。若能蒐集到相關資訊，將更

能對產業引進策略提供更完整的建議。

以DEA作策略聯盟分析之假設。分析衡量

新公司的預期經營效率之結果，為了能夠更深

入的瞭解分析對象，再運用差額變數分析得到

假設最佳效率時，應進行的投入項與產出項調

整及提出效率改善的方向。

貳、實證結果與分析

一、個案公司介紹與研究前提

G公司是一家成立二十二年的科技公司，

民國72年成立於新竹科學園區, 專門從事於

光儲存媒體的製造，目前最主要的產品為CDR

光碟片、CDRW光碟片、DVD系列以及光儲

存媒體，產品行銷世界各地，G公司主要業務

為光電產品及設備之生產與銷售，近年來隨著

相關週邊系統價格下跌與普及率提高，以及網

際網路應用頻率提高，使得需求大為增加，

2003年甚至出現供不應求的情形，有關全球光

碟片的供需情形，今年預期需求量仍持續成

長。

本論文所要討論的研究重點在於深入探討

並試著尋求光碟產業之結合模式，本文探討之

結合係指在尚未發生購併之情況下，尋求出對

購併業者與整體產業而言最佳之購併組合或

購併規則，以作為往後光碟業者在購併時之參

考依據，因此首先面臨的問題即是如何使變數

資料做「結合」

；無人能夠預知兩家甚至多家

的企業體在結合、購併後會是呈現一加一大於

一的的綜效而從此大好，亦或是面臨組織文化

之衝擊而導致經營情況每況愈下之困境，因此

本研究基於現實之考量，均假設研究資料在相

同變數間具有相加性，並以相加後之數值作為

兩家光碟業結合後之研究數據，以進行更深入

之研究。

二、研究樣本

根據

DEA 之應用程序，使用資料包絡法

的第一步驟，即是選擇適當之

DMU，本研究

之研究產業係屬電子產業中之光碟片業，因此

將挑選國內之光碟廠商為研究之標的。根據各

廠商民國

93 年之年報資料，茲將本論文之研

究光碟廠商列示於下表

1：

表

2 研究樣本總覽(指標未定)

資本額 研發經費 員工 人數 每股 盈餘 營業收入 營業淨利 錸德 21612355 548346 3548 0.64 24919369 4472010 精碟 8081993 256876 2042 2.58 9604701 1028714 中環 32490480 561827 3428 2.54 27343710 6755314 國碩 2924612 66569 453 0.79 2424211 249846 鈺德 2083953 37282 562 2.66 2810918 541457 遠茂 2055174 79157 754 7.2 2237551 789123 利碟 4020794 115098 750 0.57 4672634 158479 豐聲 365287 10417 221 2.52 601189 93857 佳錄 1571880 26388 189 6.05 259793 248147

三、投入、產出變數之選取

運用

DEA 選取變數之原則進行相關性分

分析，以驗證各變數間符合單調性質，即投入

之增加不得使產出反向減少之特性。由相關係

數分析表中，須遵守投入之增加不得使產出反

向減少之特性，發現在產出指標中，每股盈餘

與投入指標的相關係數皆為負值，而投入指標

的資本額與研發經費及員工人數跟產出指標

的營業收入及營業淨利之相關係數值皆為正

向相關，符合運用

DEA 之原則。因此刪去每

股盈餘後將投入指標的資本額與研發經費跟

產出指標的營業收入及營業淨利再做一次相

關係數分析，最終，本研究選定了以下投入、

產出項目為變數：

投入變數一

(X1)：資本額

投入變數二

(X2)：研發經費

產出變數一

(Y1)：營業收入

產出變數二

(Y2)：營業淨利

四、整合前的效率評估分級

決定了投入指標(變數)及產出指標(變

數)項目後，緊接著的效率分級工作，本研究

使用了

DEA-Solver 的軟體，採用 CCR-I 模式

做效率評估的工作，首先評估

9 家企業得到以

下的結果：

表

2 九家候選合併前效率評估分級表

Rank DMU Score

1 鈺德 1 1 遠茂 1 1 豐聲 1 4 中環 0.8279005 5 錸德 0.7723922 6 精碟 0.722086 7 利碟 0.7061118 8 佳錄 0.647497 9 國碩 0.5581941

由上表可得知，經營效率為

1 也就是經營

效率第一級的企業有｢鈺德科技、遠茂光電、

豐聲科技」三家企業，而其他企業皆未達到績

效前緣，因為不在相同效率前緣上之產業無法

相互作比較，因此接下來移除經營效率第一級

企業，再做一次經營效率評估的排名，重複使

用此種方法可以得到在相同各前緣線下各公

司之分級，如第一級前緣線效率一定比第二級

效率高，如第一級中的鈺德科技其效率就會比

第二級跟第三級中的公司來的高，而得出之結

果中各級數的產業可以相互比較，並將結果整

理如下表

3。

表

3 策略聯盟前各層級效率前緣表

第一級 達效率 前緣 鈺德科技、遠茂光 電、豐聲科技 第二級 達效率 前緣 中環、錸德科技、精 碟科技 第三級 達效率 前緣 利碟、佳錄科技、國 碩科技

五、整合過程的效率評估

此整合過程的研究階段中，必須為｢

G 公

司｣

尋找一個適宜的合併對象，於是在模擬合

併的過程裡，將精碟的投入及產出指標分別與

其他企業相加，代表兩企業的合併，當然，這

是在投入資源的合併，能得到相同產出合併的

假設前提之下，以模擬的數字，乃代表｢

G 公

司｣

與各家企業合併的假設，在假設其合併的

前提後，便是假設其為一家新企業，再與其他

原有未合併的企業做經營效率的評估了，在多

次的評估之後，尋求合併的新企業經營效率最

高且整體產業平均經營效率最高者，即為最適

當的組合了。上述模擬步驟的實際操作後，再

一次用效率分級工作，利用

DEA-Solver 的軟

體，採用

CCR-I 模式做評估的工作，評估原

先

9 家企業加上合併後的九家企業共十八家

企業後得到以下的結果，但是依照

DEA 原

理，每個非效率前緣

DMU 比較單位都是最接

近之效率前緣

DMU，而不同參考 DMU 其

SCORE 是無法比較的，因此非效率前緣 DMU

是無法比較的，而本研究的重點是對於需要合

作的企業是否有效率提升的現象，因此將第一

級效率前緣

DMU 移除後，再透過評估了解第

二級

DMU 效率前緣單位，如此類推，評估至

達

DMU 數目達最低表準時，經過整理結論如

下表

4:

表

4 策略聯盟後各層級效率前緣表

第一級 鈺德、豐聲、鈺德+G 公司、豐聲+G 公司 第二級 錸德+G 公司、中環+G 公司、國碩+G 公司、錸德 遠茂+G 公司、利碟+G 公司、佳錄+G 公司、中環 第三級 精碟+G 公司、遠茂、利碟 第四級 精碟、佳錄 第五級 剩餘企業 國碩

根據以上資料整理出策略聯盟推薦表如

表

5，在差異欄位中，左邊值代表原企業在策

略聯盟前的分級數，而後面值代表策略聯盟後

評估出來的分級數，經過差異可以了解其聯盟

後是否效率能提升績效，也是本研究欲分析的

結果，而數值差異

0 為一般，負值代表越差，

相對正值越大代表愈值得推薦。以遠茂+G 公

司為例，原本公司評估為第三級企業，在合作

聯盟後變成第二級企業因此差異為

3-2=1，因

此在推薦表中為好因此推薦，差異

0 為一般，

負值代表越差，相對正值越大代表愈值得推

薦。

其實並不一定要跟該產業中的最佳的公

司作策略聯盟才能提升小廠商的競爭力，因為

小公司必須有足夠吸引大公司的能力大公司

才會願意雨小公司作策略聯盟，所以小公司其

實也可與產業中其他小型廠商作策略聯盟，進

而提升競爭力，來與產業中最佳的廠商相抗

衡，而下表就是經過分析後的策略聯盟推薦

表，讓產業在選擇策略聯盟的伙伴時能有更好

的參考資料，以瞭解到底跟誰策略聯盟才能達

到雙贏的局面。

表

5 策略聯盟推薦表

Rank company 差異 最佳 佳 好 一 般 差 較差 極差 1 鈺德 1 豐聲 1 鈺德+G 公司 1>1 0 1 豐聲+G 公司 1>1 0 2 錸德+G 公司 2>2 0 2 中環+G 公司 2>2 0 2 國碩+G 公司 2>2 0 2 遠茂+G 公司 2>3 1 2 利碟+G 公司 2>3 1 2 佳錄+G 公司 2>4 2 2 錸德 2 中環 3 精碟+G 公司 3>4 1 3 遠茂 3 利碟 4 精碟 4 佳錄 5 國碩

六、差額變數分析

我們運用了差額變數之分析，針對資源配

置不當之變數提供改善之依據，並分析資源分

配之過多或不足的現象是否有出現在投入與

產出變數項目間，因此於本節中，以利碟與

G

公司聯盟為例，將使用差額變數分析之方法，

瞭解其資源應用之情況，並予以分析結合之結

果。

表

6 利碟與 G 公司策略聯盟後之差額變數

分析

DMU Score

I/O Data Projection Difference % 利碟+G 公司 0.88667 資本額 7961633 7059356 -902276 -11.33% 研發經費 387720 343780 -43939 -11.33% 營業收入 15304080 15304080 0 0.00% 營業淨利 3496014 3880930 384916 11.01%

根據表

6 之差額變數分析，如表中的第 1

組結合：

（利碟+G 公司），本研究可提供未達

高效之結合組合一改善之方向，利碟+G 公司

其策略聯盟後之效率值為

0.886671986，並未

達高效之標準，因此可由差額變數分析來瞭解

其未達高效之原因，以此例而言，可看出策略

聯盟後需要在投入及產出的各項變數上予以

增加或縮減，才能達到效率前緣值為

1 的目

標，而縮減之幅度則分別為投入變數的資本額

11.33%、研發經費 11.33%；增加之幅度為產

出變數的營業淨利

11.01％；而營業收入因為

變動幅度顯示為

0.00%所以不需變動。

參、結論與建議

本研究選定兩項投入指標(資本額及研發

經費)與兩項產出指標(營業收入及營業淨利)

以評估台灣光碟產業

9 家企業的經營效率，接

著找出合併後新企業經營效率最高者而且整

體產業平均經營效率最高者，此即為最適當的

組合了。最後經歷評估合併工作後，有

2 家整

體平均經營效率為目標值

1，而有 7 家的整體

平均經營效率為進步的，結果證實，此研究結

果可讓正在考慮是否要採取策略聯盟行動的

廠商，能有更進一步的評估工具，而策略聯盟

並不一定能增加績效，如果盲目地進行策略聯

盟，有時對企業反而會造成更大的損害，所以

有效的策略聯盟更可對企業在策略聯盟時提

出建議改善方案，以利產業評估策略聯盟選擇

對象時，選擇符合合作夥伴之策略，為企業在

進行策略聯盟時，建構有利產業發展之評估工

具及架構。

因此本論文主要貢獻整理如下(1)過去資

料包絡法的論文多半僅止於績效評估而已，以

策略聯盟為主的論文則為少數，故本論文可以

為後來欲研究此方向的論文，作為一個起始，

(2) 本文針對相同產業、同業間之策略聯盟模

式，提供了一研究之依據，並就本文所探討之

產業，提供一策略聯盟之參考策略。(3)為正

在考慮是否要採取策略聯盟行動的廠商，提供

一評估 工具

(4) 運用差額變數分析效率發

現，廠商讓總效率提升的方法，並不是只有一

昧的增加投資，有時在進行策略聯盟時需要增

加或減少投入，也能達到同樣的效果(5) 一般

DEA 文獻皆針對 SCORE 做績效評估，但是本

人認為在不同參考點的情況下，對於全部

SCORE 做討論，難免會落入誤差的迷失，因

此本研究提出的

RANK 排序法，能確切得比

較出受評單位的排序優劣

本 研 究資 料 基 於研 究之 限 制， 僅選 取

2003 年產業之相關資訊以為研究，若能在研

究產業之資料蒐集上兼具深度與廣度，相信必

能有所發現與突破，另外本研究僅針對一個產

業作經營效率評估，在未來後續研究中可以加

入對於水平整合或是上游及下游協力廠商之

策略聯盟探討以增加研究之主題，另一方面也

可在產出變數與投入變數作更廣泛與深入的

蒐集，凡對於績效有所影響之指標，如生產產

值、股東權益…等，均可納入研究變數中，建

議深入瞭解相同產業之同業間的優劣關鍵，可

採取更貼近研究所屬產業全貌與多元之變數

項目予以探討，進而提供組織於管理上更積極

有利的管理資訊，以及探討評估是否該縮減企

業，也就是所謂的企業瘦身，探討要如何達到

正確的企業縮減，這些都是未來相當具備前瞻

性的研究主題。

四、參考文獻

1. Banker, R.D, A. Chanes and W.W. Cooper. (1984).

Some Models for Estimating Technacal and

Scale Inefficiencies in Data Envelopment

Analysis, Management Science, 30: 1078-1092.

2. Bhattacharyya A., Lovell, C. A. K. and Sahay, P.

(1997). The Impact of Liberalization on the

Productive Efficiency of Indian Commerical

Banks. Europen Journal of Operatiojnal Research,

98: 332-345.

3. Charnes A., Cooper, W. W. and Rhodes, E. (1978).

Measuring the Efficiency of Decision Making

Units.

European

Journal

of

Operational

Research, 2 (6): 429-444.

4. Charnes, A., W. W. Cooper, L. Seiford and J.

Stutz. (1983). Invariant multiplicative efficiency

and piecewise Cobb-Douglas envelopment.

Operations Research Letters, 2(3): 101-103

5. Charnes, A., W. W. Cooper, L., B. Golany, L.

Seiford and J. Stutz. (1985). Foundations of data

envelopment

analysis

for

Pareto-koopmans

efficient empirical production functions. Journal of

Econometrics, 30: 91-107.

6. Farrell, M.J., M. Fieldhouse. (1962). Estimating

efficient production under increasing returns to

scale. Journal of the Royal Statistical Society,

Series A, 125: 252-267.

7. Farrell,M.J. (1957). The Measurement of

Productive Efficiency. Journal of the Royal

Statistical Society, Series A, General 120: 253-281

8. Jemric, Igor and Vujcic, Boris. (2002). Efficiency

of Banks in Croatia- A DEA Appr

oa

c

h”

,

Comparative Economic Studies, XLIV, No.2

(summer): 169-193.

Evaluate the potential suppliers’ performance based on heuristic algorithm and

Malmquist model: cased by IC packaging and testing industry of Taiwan

Chia-Nan Wang, Hsun-Cheng Liu

*

Department of Industrial Engineering and Management

National Kaohsiung University of Applied Sciences

E-mails:

*

[email protected]

Abstract

To enhance the performance of global supply chain is one of the important competitive strategies for many businesses. The performance evaluation of suppliers within chain has become an important subject for impacting the efficiency of chain. It is not realistic to evaluate the performance by just one short period of time. To integrate the data across years would be a better way. Moreover, the suppliers always have different resources base, such as different inputs and outputs. Therefore, the concept of efficiency of data envelop analysis (DEA) is applied for this study. This study develops a systematic method based on the combination of the Malmquist productivity index model in DEA and heuristic algorithms to evaluate performances for different based suppliers within a period of several years. The case study of 14 companies in the IC packaging and testing industry listed in Taiwan’s Stock exchange are applied to verify the feasibility of proposed method. The results show that two firms, Advanced Semiconductor Engineering Inc. and Phoenix Precision Technology, would be better potential suppliers. Therefore, this study can be used to recommend better suppliers for better efficiency performance under different base resources and cross years time frame.

1. INTRODUTION

The supply chain management has become an essential trend for gaining competitive advantage. Assembling, packaging, and testing of semiconductors, these words signal a huge amount of capital expenditure [1]. However, today’s world-wide market has become a knowledge economy. The trend that the IC packaging and testing industry continues to advance toward the high-end packaging and testing services has also gradually pushed the industry toward becoming a high capital investment industry. The booming of Taiwan's semiconductor industry has created Taiwan's economic prosperity for the last 10 years. Professional outsourcing for each key process is the critical factor for the success of Taiwan's semiconductor industry.

From design, manufacturing, packaging, and testing, especially in the backend processing of the

packaging and testing industry, a number of manufacturers were set up around 1990 and formed a highly competitive scene. Chang and Tsai [2] propose that because Taiwan's IC industry can not quickly attain the leading position in technology, it should focus on following the leaders as rapidly as possible. Thus, there is not only a continuous need to expand their equipment and machinery, but also a need for vertical and horizontal integration etc. to increase the manufacturing capacity. M & A was a subject often discussed and manipulated by the industry in recent years. The efficient cooperation of supplier and customer will enhance competitive. Therefore, supply chain management can help in achieving the both goals for of supplier and customer in this industry simultaneously.

Take Taiwan as an example: In the past, Taiwan’s enterprises rose prominently in the international market place with a posture of “small is beautiful.” At its peak time, Taiwan owned more than 40 IC packaging and testing manufacturers at the same time. Price competition has been very severe. And now, the high operation risk accompanying the high capital investment has indeed built a stiff entry barrier for the IC packaging and testing industry. However, it has created a positive influence to the long-term development of the IC packaging and testing industry. Actually, the leading companies among the Taiwan IC packaging and testing companies wield a huge influence in Taiwan and even in the global market [3]. Therefore, to enable Taiwan's semiconductor industry to transit into a high-value-added industry more concentrating on design innovation is one of the main focuses that the industry and Government should be concerned and converging on. However, it is still an imperative issue to improve competitive advantage for the IC industry; in particular, the R & D technology and productivity performance of the IC industry is of great importance.

From the point of view of vertical and horizontal integration: the results of these studies helped achieving economic integration, product differentiation, cost reduction, the ability to obtain market information, raising the entry barrier, the driving force for growth, expansion of product and market scale, and market prices influence etc. However, the disadvantages were that internal control and external coordination prone to problems,

encountering blockade and crowded-out effect, and managing diverse risks etc. The importance of an adequate framework for the selection of suppliers, therefore selection of appropriate supplier is one of the fundamental strategies. In order to solve above problems, combined with Data envelopment analysis (DEA) and heuristic method, this study proposes an approach to evaluate the efficiency of groups of company. One of the DEA models, which is called Malmquist productivity index (MPI), is applied. DEA method has advantage of analyzing multiple inputs and multiple outputs against many targeted organizations. Therefore this research will utilize DEA to assess the empirical relationship between inputs and outputs for the targeted assembly housed in Taiwan for the past four years.

This paper is organized as follows: Section 2 reviews the previous literature on the topic. This proposed approach is applied to evaluating the IC packaging and testing industry of Taiwan in Section 3. Section 4 demonstrates the experiment with realistic data and analysis. Concluding remarks is presented Section 5.

2. Literature review

In the past two decades, more than 3,000 publications have adopted DEA and/or MPI as the evaluation techniques to assess firm efficiency. Gulati [4] stated that despite the extensive research on collaborative relationships and networks, but our understanding of the potentially adverse effects of networking is relatively limited. Bhutta and Huq [5] pointed out the supplier chain problem requires the consideration of multiple objectives, and hence can be viewed as a multi-criteria decision-making (MCDM) problem. Many more procedures and methods, Weber [6] suggested the use of Data envelopment analysis (DEA) for measuring vender performance in the first issue of this journal. Statistical approaches have been examined Δ Narasimhan [7] described analytical hierarchical processing approach. Leenders et al. [8] suggested including simple weighted rating, AHP, multi-attribute utility theory, mathematical programming, game theory, principal components analysis and neural networks. The following are broader use of DEA Malmquist: Kim and Lee [9] applied DEA Malmquist to measure the spillover of manufacturing technology R & D; Wu et al. [10] used DEA Malmquist for Taiwan’s IC design companies; Chen and Ali [11] applied DEA Malmquist to measure computer industry; Liu and Wang [12] used DEA Malmquist to measure Taiwan semiconductor companies.

Based on the prior studies on the selection of the input measures indexes include: Number of employees [10], Labor costs [13], fixed capital [9],

R&D expenditures [10], Market condition [14]; output measures indexes include: Net sales [10], Patents [15], Value added [9]. Patents have face validity as measures of technological knowledge [16]. Patents provide an index of a firm’s variety-generating capability, thus represent an alternative to R&D expenditures or number of R&D employees for measuring firm innovation capacity.

3. The proposed methodology

DEA was first introduced by Charnes et al. [17]. They proposed the CCR model to measure the maximum possible relative efficiency of Decision Making Units (DMUs) in 1978. DEA is different from using the pre-determined production function in commonly used statistical methods. There is no prior information for the trade-off relationship between inputs and outputs in DEA. It is a non-parametric method and uses mathematical programming to estimate the efficiency frontier based on the concept of relative comparison to measure the efficiency value of multiple input indexes and multiple output indexes in the DMUs. Assuming there are n DMUs and x andij

y

rjrepresent the value of the ith input

index(i=1,2,...,m)of the jth DMUj(j=1,2,...,n) and the value of the rth output index(r=1,2,...,s). The difference variables of the ith input index and the rth output index are represented by si andsr+;

they mean the amount to be subtracted from input and the amount to be added to output for the efficiency improvement respectively. Variable Ȝ is j

the reference weight of DMU to 0 DMU . It goes j

even further than replacing the CCR model with the SBM model and uses four distance functions in the Malmquist index calculation. The Malmquist productivity index was proposed by Caves et al. [18]. If D=

(

x0,y0

)

=1 ,

DMU

o (Object) is high

efficiency and is not low efficiency. Fare et al. [19] partition the Malmquist productivity into two parts: one part is the changes of the catching-up inefficiency and the other part is the frontier efficiency. Here _CIEtШt+1 _{is defined as the}

catching-up inefficiency from t period to t+1 period:

1 t t

CIEo is basically the geometric mean of the two measurements. t t t t t t t t t t OZ OZ OZ OZ TSE TSE CIE 1 1 1 1 1      o ₍₁₎

unit measured at t period, that is the distance moved by the projected t

t

Z+1 at t+1 period. The later

measures the projected Ztt1 for the Zt1 unit

measured at t period, that is the distance moved by the projected +1

1 +

t t

Z at t+1 period. The details are shown as follows: 2 1 1 1 1 1 1 1 1 » » » » ¼ º « « « « ¬ ª u u        t t t t t t t t t t t t t t OZ OZ OZ OZ OZ OZ OZ OZ MI (2)

Combine(1)and(2), the Malmquist IndexΰMPIαis shown in 1 1 1 o   _u t t t t t t CIE MI MPI 2 1 1 1 2 1 1 1 » » ¼ º « « ¬ ª » » ¼ º « « ¬ ª  o o    t t t t t t t t IEI IEI TSE TSE TSE TSE 2 1 1 1 2 1 1 » » ¼ º « « ¬ ª » » ¼ º « « ¬ ª  o o   t t t t t t IEI IEI TSE TSE (3) 1 + t t

MPI >1 means that there is significant increase in the productivity,MPItt1<1 means productivity has

decreased and MPItt1=1 means that there is no

change in productivity. Therefore, considering of these ways, we apply the Malmquist model to evaluate the efficiency.

4. Research design and Methodology

In recent years, as the competition in the IC packaging and testing industry has become more and more intensive. This most products of this industry are short lifting cycles. Accordingly, the supply chain impact will be important factor of competition. It is very important to choose appropriate input and output terms in applying DEA. In addition, when choosing and assessing decision-making units, it is necessary to consider whether the goals for operation are similar, whether they are operating under the same market conditions, and whether the input, output factors for performance characteristics shown by each unit are the same. If performance assessments are carried out under different scenarios, the measured results will be biased. In this chapter, the samples and data for variables chosen are analyzed using SPSS packaging software to assess the correlation between the input and output variables and whether they show isotonicity for Taiwan’s 14 IC packaging and testing companies.

DEA method is deployed to calculate the input direction Malmquist index of DMUs for 14 IC packaging and testing companies and to analyze the efficiency for each technology and the productivity index. These results serve as references for management and policy-makers to address the shortcomings of integration so that they can achieve an efficient operation in the future to enhance competitiveness. The proposed research process for this study is shown as follows:

Figure1. Research development Step1. Choosing the Decision-making unit

Matching the scope and the target of the research subject, 14 qualified companies are selected from companies listed on the Taiwan’s stock market with business areas crossing IC packaging and testing. The selection of the candidate companies are based on the criteria that they are having business, having the same portion of the relevant sales channels, having some kind of relationship for initial cooperation, or even competitor with the company in case.

Step2.Choosing the input and output variables Literatures that assess the efficiency of the same

Malmquist analysis Frontier-shift analysis Catch-up analysis Relationship Ї0.05

Step4.Data envelopment analysis Step1.Data acquisition

Step2. Discrimination factors

Calculate the relative interaction of factors Step3.Statistical analysis

Step5.Rank

Step6. Results analysis

Step7. Conclusion NO

kind of high technology industries as this study are used to find the variables that are suitable for this study.

Step3. Related analysis

In order to find out whether the input and output factors that are suggested by the literature review are highly correlated, the input and output variables are analyzed for correlation using statistical models to see whether the significant level of the two-tailed test is larger than 0.05. If it is larger than 0.05, they satisfy the isotonicity requirement. Otherwise, we go back to step 1 and choose new Decision-making unit again until it meets the criteria.

Step4. Index analysis

After the high correlation of the input and output are determined for the 14 companies in this study, these data are manipulated using DEA model. Malmquist index is further used with the multi-phase, input-oriented, efficient frontier data to assess their performance, analyze their technical efficiency, analyze their position, and recommend the integration solution for each company in case.

Step5.Sorting of the Decision-making units

After using DEA Malmquist index to assess the performance and the technical efficiency of the 14 companies, these companies are ranked by their rate of productivity progress and changes in the level of technical efficiency.

Step6. Results analysis

This section summarizes the ranking results for the Decision-making units and makes recommendations on how to integrate with the suppliers, fend off the external competitors, and utilize explicit and implicit knowledge to correct the shortcomings of horizontal and vertical integration so that managers can enhance the competitiveness of the enterprise.

Step7.Conclusions and recommendations

Suitable recommendations are given and conclusions and future research directions are summarized based on the selection and filtering of the factors that impact suppliers, the use of DEA model to assess the productivity and the technical efficiency for each supplier, and the ranking of the suppliers and the result analysis.

5. Empirical analysis and Results

5.1 Collect the DMU

The targets of this study are 14 domestic listed major IC packaging and testing companies. They are summarized in Table1. The data used are 2004, 2005, 2006, and 2007 related data for each packaging and testing company. The data are collected from the following sources:(1)Taiwan’s patent bulletin (2)public information observation websites (3)Taiwan’s Stock Exchange website.

Table 1: Company list

5.2 Input/Output selection

Based on the literature review mentioned previously and the elements of the operation for IC packaging and testing, the indicators chosen for input measurements in this study include: R&D expenditure Ε Labor costs Ε Capital Ε Number of employeesΕFixed capitalΕsample of R&D projectsΕ Market condition and the indicators for the output include: Sales revenuesΕTechnology improvementΕ Value addedΕ Net salesΕPatentsΕ NPV. Due to the limitation by the access to data in this study, the aforementioned input and output indicators can not be defined and measured based on quantified data. Furthermore, these 14 companies do not provide needed data for us. Therefore, this study can only select 4 inputs and 2 outputs for the empirical analysis.

This study uses 4 input items and 2 output items. The definition of the input and output variables, the input and output items are divided based on the literature review mentioned. The input and output variables are calculated to see whether their correlation is of isotonicity. If anyone of the three rules is violated, we will go back to step 1 of the selection process to re-do the variable selection until all rules are satisfied. Because DEA can not handle properly the case that the output is negative, that is, if input increases, output should not decrease. The input and output items are checked using the Person correlation method to see what kind of relation between the input and the output (positive or negative relationship). Table2 shows that the input variables are positively correlated and the input and the output are positively correlated fairly significantly. Therefore, the input and output items are all of the isotonicity nature.

Code IC packaging and testing company F1 LINGSEN PRECISION INDUSTRIES, LTD.

F2 ADVANCED SEMICONDUTOR ENGINEERING, INC. F3 GREATEK ELECTRONTCS INC.

F4 CHIPBOND TECHNOLOGY CORPORATION. F5 SILICONWARE PRECISION INDUSTRIES CO.,LTD. F6 FUPO ELECTRONICS CORPORATION. F7 ARDENTEC CORPORATION. F8 POWERTECH TECHNOLOGY,INC. F9 INTERNATIONAL SEMICONDDUCTOR

TECHNOLOGY, LTD.

F10 SHENMAO TECHNOLOGY, INC. F11 PHOENIX PRECISION TECHNOLOGY

CORPORATION

F12 KING YUAN ELECTRONICS CO., LTD. F13 SIGURD MICROELECTRONICS CORP. F14 JOYIN CO;LTD

Table2: Person correlation coefficients

Table2 lists the Person correlation coefficients. From the table, it can be seen that although all the factors are positively correlated, the correlation between Employees and Fixed capital input is only 0.272. It is obviously much lower than the positive correlation among other factors but does not violate the DEA model, which can not handle negative correlation. We further investigate whether there is isotonicity between the input and the output variables.

Table3: Technical efficiency, technical level, and productivity Index DMUs Catch-up 2004=>2007 Frontier 2004=>2007 Malmquist 2004=>2007 F1 _{0.762565 1.181872 0.901254} F2 _{1.114991 2.306376 2.571588} F3 _{1.398262 0.870487 1.217169} F4 _{0.756149 1.105831 0.836173} F5 _{0.803607 1.405983 1.129858} F6 _{0.822162 1.374351 1.12994} F7 _{0.829201 1.12765 0.935048} F8 _{0.629664 1.177781 0.741607} F9 _{0.522803 1.512709 0.790849} F10 _{1.209485 1.218146 1.473329} F11 _{0.727563 2.74978 2.000638} F12 _{0.909162 1.072031 0.97465} F13 _{1.225079 1.486232 1.820751} F14 _{0.583761 1.84552 1.077343} Average 0.878175 1.459625 1.257157

From Table3, we know that, from 2004 to 2007, F2,F3, F5, F6, F10, F11,F13, and F14, these 8 companies exhibit the best scenario for productivity improvement. The reason for the improvement of the productivity is due to the combination of technical ability moving in the positive direction and the improvement of the relative efficiency. From 2004 to 2007, F1, F4, F7, F8, F9, andF12, these 6 companies exhibit the worst scenario for the declining productivity. The reason for the declining of the productivity is due to the combination of technical

ability moving in the negative direction and the declining of the relative efficiency.

6. Concluding remarks

IC packaging and testing industry is the future star industry in the semiconductor industry. The practices that many enterprises use to obtain a competitive advantage, control market channels, and bring down competitors through the uses of merger and acquisitions for each other and strategic alliance etc. to achieve business integration have become a new trend of business management. How to choose suitable suppliers is a very important issue to any organization in a highly competitive era of global change. There are many factors need to be considered, which include quantitative and qualitative factors. Occasionally, it is necessary to obtain the performance of a company at some critical moments by using the elements that constitute the Malmquist index to analyze each factor and information to explore the context of the business management. In addition, the trend is to use DEA to obtain Malmquist indexes efficiently. It is very helpful in the analysis of organizations in each industry.

In terms of choosing suppliers, this study uses DEA Malmquist model analysis to conduct performance assessment of empirical research, to measure the combination of financial factors as quantitative indicators and patents and the technical level as qualitative indicators, and to analyze the operating performance of suppliers for 14 Taiwan Semiconductor packaging and testing companies. By using DEA Malmquist model analysis, it is not only helpful in finding shortcomings for the technical level but also decision-makers can take into account the technical capacity of suppliers at the same time as a way to be considered for the supply chain integration. They will no longer be restricted by the customer and supplier relationship and can enhance partnership for cooperation to achieve an efficient operation in the future so that their competitiveness can be enhanced.

Because F2 and F11 focus on the IC Carrier and high end packaging business, if they want to expand the market oversea and provide Turn key packaging and testing services, it is necessary to acquire the low end part of the packaging and testing. They can consider forming alliance with F3 to get the necessary technical exchange and experience. If they want to cross over to memory test such as logic IC and mixed-signal IC, they can consider forming alliance with F5, F9 and F8. It will not only enhance the technical performance but also will take into account the further expansion of market and the issues of acquiring new technology. In terms of DEA efficiency, overall technical efficiency, pure technical efficiency, and scale efficiency provide objective

R&D expenditure Employees Labor costs Fixed capital Patents Net sales R&D expenditure 1 Employees 0.318 1 Labor costs 0.816 0.406 1 Fixed capital 0.690 0.272 0.908 1 Patents 0.725 0.624 0.814 0.652 1 Net Sales 0.799 0.406 0.985 0.901 0.780 1

standard for comparison. They allow the Decision-making units in the semiconductor industry in this study, from the viewpoint of productivity, to conduct cross-year and cross-unit comparison comprehensively and provide benchmark for other Decision-making units.

7. References

[1] Che, Z. H., Wang, H. S., and Sha, D. Y., “A multi-criterion interaction-oriented model with proportional rule for designing supply chain networks”, Expert Systems with Applications, Vol. 33, pp. 1042-1053, 2007.

[2] Chang, P. L., and Tsai, C. T., “Finding the niche position-competition strategy of Taiwan’s IC design industry”, Technovation, Vol. 22, No. 2, pp. 101-111, 2002.

[3] Wang, P. H., and Liu, F. H., “DEA Malmquist productivity measure: Taiwanese semiconductor companies”, International Journal of Production

Economics, Vol. 112, pp. 367-379, 2008.

[4] Gulati, R., “Alliances and networks”, Strategic

Management Journal, Vol. 19, No. 4, pp. 293–317,

1998.

[5] Bhutta, K. S., and Huq, F., “Supplier selection problem: a comparison of the total cost of ownership and analytic hierarchy process approaches”, Supply

Chain Management: An International Journal, Vol. 7,

No. 3, pp. 126-35, 2002.

[6] Weber, C. A., “A data envelopment analysis approach to measuring vendor performance”, Supply Chain

Management, Vol. 1, No. 1, pp. 28-39, 1996.

[7] Narasimhan, R., “An analytical approach to supplier selection”, Journal of Purchasing and Materials

Management, pp. 27-32, 1983.

[8] Leenders, M., Johnson, P. F., Flynn, A., and Fearon, H. E., Purchasing Supply Management, 13th_{, New} York: McGrawHill, 2006.

[9] Kim, K. J., and Lee, H. K., “Embodied and disembodied international spillovers of R&D in OECD manufacturing industries”, Technovation, Vol. 24, pp. 359-368, 2004.

[10] Wu, W. Y., Tsai, H. J., and Cheng, K. Y., “Assessment of intellectual capital management in Taiwanese IC design companies: using DEA and the Malmquist productivity index”, R&D Management, Vol. 36, No. 5, pp. 531-545, 2006.

[11] Chen, Y., and Ali, A. I., “DEA Malmquist productivity measure: new insights with an application to computer industry”, European Journal

of Operational Research, Vol. 159, pp. 239–249,

2004.

[12] Franklin Liu, F. H., and Wang, P. H., “DEA Malmquist productivity measure: Taiwanese semiconductor companies”, International Journal of

Production Economics, Vol. 112, pp. 367–379, 2008.

[13] Gonzalez, E., and Gascon, F., “Sources of productivity growth in the Spanish pharmaceutical industry (1994–2000)”, Research Policy, Vol. 33, pp. 735-745, 2004.

[14] Sohn, S. H., and Moon, T. H., “Decision Tree based on data envelopment analysis for effective technology commercialization”, Expert Systems with Applications,

Vol. 26, pp. 279-284, 2004.

[15] Wang, E. C., and Huang, W., “Relative efficiency of R&D activities: A cross-country study accounting for environmental factors in the DEA approach”,

Research Policy, Vol. 36, pp. 260-273, 2007.

[16] SubbaNarasimha, P. N., Ahmad, S., and Mallya, S. N., “Technological knowledge and firm performance of pharmaceutical firms”, Journal of Intellectual Capital, Vol. 4, No. 1, pp. 20–33, 2003.

[17] Charnes, A. A., Cooper, W. W., and Rhodes, E., “Measuring the efficiency of decision making units”,

European Journal of Operational Research, Vol. 2,

pp. 429-444, 1978.

[18] Cave, D. W., Christensen, L. R., and Diewert, W. E., “The Economic Theory of Index Numbers and the Measurement of Input Output, and Productivity”,

Econometrica, Vol. 50, pp. 1393-1414, 1982.

[19] Fare, R. S., Grosskopf, S., Lindgren, B., and Roos, P., “Productivity Changs in Swedish Pharmacies 1980-1989: A Non-Parametric Malmquist Approach,”

Journal of Productivity Analysis, Vol.3, pp.85-101,