Coalition Formation and Data Envelopment Analysis

Similar documents
CHAPTER 3 INEQUALITIES. Copyright -The Institute of Chartered Accountants of India

Pattern Recognition 2014 Support Vector Machines

Computational modeling techniques

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank

Differentiation Applications 1: Related Rates

Least Squares Optimal Filtering with Multirate Observations

WRITING THE REPORT. Organizing the report. Title Page. Table of Contents

Floating Point Method for Solving Transportation. Problems with Additional Constraints

Thermodynamics Partial Outline of Topics

A New Evaluation Measure. J. Joiner and L. Werner. The problems of evaluation and the needed criteria of evaluation

Department of Economics, University of California, Davis Ecn 200C Micro Theory Professor Giacomo Bonanno. Insurance Markets

Math Foundations 20 Work Plan

COMP 551 Applied Machine Learning Lecture 5: Generative models for linear classification

CS 477/677 Analysis of Algorithms Fall 2007 Dr. George Bebis Course Project Due Date: 11/29/2007

Chapter 5: The Keynesian System (I): The Role of Aggregate Demand

Inflow Control on Expressway Considering Traffic Equilibria

Bootstrap Method > # Purpose: understand how bootstrap method works > obs=c(11.96, 5.03, 67.40, 16.07, 31.50, 7.73, 11.10, 22.38) > n=length(obs) >

x 1 Outline IAML: Logistic Regression Decision Boundaries Example Data

Support-Vector Machines

The blessing of dimensionality for kernel methods

COMP 551 Applied Machine Learning Lecture 11: Support Vector Machines

Fall 2013 Physics 172 Recitation 3 Momentum and Springs

Land Information New Zealand Topographic Strategy DRAFT (for discussion)

A Matrix Representation of Panel Data

Section 6-2: Simplex Method: Maximization with Problem Constraints of the Form ~

Optimization Programming Problems For Control And Management Of Bacterial Disease With Two Stage Growth/Spread Among Plants

SUPPLEMENTARY MATERIAL GaGa: a simple and flexible hierarchical model for microarray data analysis

DEA Models for Two-Stage Processes: Game Approach and Efficiency Decomposition

IAML: Support Vector Machines

Preparation work for A2 Mathematics [2017]

Biplots in Practice MICHAEL GREENACRE. Professor of Statistics at the Pompeu Fabra University. Chapter 13 Offprint

INSTRUMENTAL VARIABLES

LHS Mathematics Department Honors Pre-Calculus Final Exam 2002 Answers

Introduction: A Generalized approach for computing the trajectories associated with the Newtonian N Body Problem

Determining Optimum Path in Synthesis of Organic Compounds using Branch and Bound Algorithm

Administrativia. Assignment 1 due thursday 9/23/2004 BEFORE midnight. Midterm exam 10/07/2003 in class. CS 460, Sessions 8-9 1

A Few Basic Facts About Isothermal Mass Transfer in a Binary Mixture

Lecture 13: Electrochemical Equilibria

Technical Bulletin. Generation Interconnection Procedures. Revisions to Cluster 4, Phase 1 Study Methodology

Revision: August 19, E Main Suite D Pullman, WA (509) Voice and Fax

Analysis on the Stability of Reservoir Soil Slope Based on Fuzzy Artificial Neural Network

Evaluating enterprise support: state of the art and future challenges. Dirk Czarnitzki KU Leuven, Belgium, and ZEW Mannheim, Germany

Enhancing Performance of MLP/RBF Neural Classifiers via an Multivariate Data Distribution Scheme

Lesson Plan. Recode: They will do a graphic organizer to sequence the steps of scientific method.

Chapter 3: Cluster Analysis

NGSS High School Physics Domain Model

Product authorisation in case of in situ generation

ANSWER KEY FOR MATH 10 SAMPLE EXAMINATION. Instructions: If asked to label the axes please use real world (contextual) labels

DEA models with production trade-offs and weight restrictions

A Quick Overview of the. Framework for K 12 Science Education

4th Indian Institute of Astrophysics - PennState Astrostatistics School July, 2013 Vainu Bappu Observatory, Kavalur. Correlation and Regression

Physics 2B Chapter 23 Notes - Faraday s Law & Inductors Spring 2018

Pressure And Entropy Variations Across The Weak Shock Wave Due To Viscosity Effects

Computational modeling techniques

What is Statistical Learning?

How do scientists measure trees? What is DBH?

DECISION BASED COLLABORATIVE OPTIMIZATION

ChE 471: LECTURE 4 Fall 2003

We can see from the graph above that the intersection is, i.e., [ ).

Hypothesis Tests for One Population Mean

Department of Electrical Engineering, University of Waterloo. Introduction

Computational modeling techniques

NUROP CONGRESS PAPER CHINESE PINYIN TO CHINESE CHARACTER CONVERSION

Lead/Lag Compensator Frequency Domain Properties and Design Methods

The standards are taught in the following sequence.

Accreditation Information

On Boussinesq's problem

Building Consensus The Art of Getting to Yes

ASSESSMENT OF REGIONAL EFFICIENCY IN CROATIA USING DATA ENVELOPMENT ANALYSIS

MATHEMATICS SYLLABUS SECONDARY 5th YEAR

Lecture 17: Free Energy of Multi-phase Solutions at Equilibrium

x x

Simulation of the Coating Process

Admissibility Conditions and Asymptotic Behavior of Strongly Regular Graphs

Green economic transformation in Europe: territorial performance, potentials and implications

Engineering Decision Methods

Computational modeling techniques

LECTURE NOTES. Chapter 3: Classical Macroeconomics: Output and Employment. 1. The starting point

Lim f (x) e. Find the largest possible domain and its discontinuity points. Why is it discontinuous at those points (if any)?

3. Mass Transfer with Chemical Reaction

Competency Statements for Wm. E. Hay Mathematics for grades 7 through 12:

Preparation work for A2 Mathematics [2018]

Writing Guidelines. (Updated: November 25, 2009) Forwards

Comprehensive Exam Guidelines Department of Chemical and Biomolecular Engineering, Ohio University

making triangle (ie same reference angle) ). This is a standard form that will allow us all to have the X= y=

Reinforcement Learning" CMPSCI 383 Nov 29, 2011!

8 th Grade Math: Pre-Algebra

The Kullback-Leibler Kernel as a Framework for Discriminant and Localized Representations for Visual Recognition

You need to be able to define the following terms and answer basic questions about them:

MATHEMATICS Higher Grade - Paper I

MODULE FOUR. This module addresses functions. SC Academic Elementary Algebra Standards:

APPLICATION OF THE BRATSETH SCHEME FOR HIGH LATITUDE INTERMITTENT DATA ASSIMILATION USING THE PSU/NCAR MM5 MESOSCALE MODEL

Tree Structured Classifier

In the OLG model, agents live for two periods. they work and divide their labour income between consumption and

February 28, 2013 COMMENTS ON DIFFUSION, DIFFUSIVITY AND DERIVATION OF HYPERBOLIC EQUATIONS DESCRIBING THE DIFFUSION PHENOMENA

STATS216v Introduction to Statistical Learning Stanford University, Summer Practice Final (Solutions) Duration: 3 hours

Chapter 2 GAUSS LAW Recommended Problems:

Module 4: General Formulation of Electric Circuit Theory

CONSTRUCTING STATECHART DIAGRAMS

Pipetting 101 Developed by BSU CityLab

Transcription:

Jurnal f CENTRU Cathedra Vlume 4, Issue 2, 20 26-223 JCC Jurnal f CENTRU Cathedra Calitin Frmatin and Data Envelpment Analysis Rlf Färe Oregn State University, Crvallis, OR, USA Shawna Grsspf Oregn State University, Crvallis, OR, USA Dimitris argaritis * University f Aucland Business Schl, Aucland, New Zealand Abstract The intrductin f a framewr fr ptimal calitin frmatin using data envelpment analysis (DEA) methds is the fcus f this paper. Simple eamples illustrate hw DEA is useful in frmulating calitin mdels and deriving ptimal slutins. In particular, the paper shws the relevance f the prpsed framewr in the cntet f analyzing hw cmpanies may reach decisins t acquire ptential partners. eywrds: DEA (data envelpment analysis), calitin frmatin, cre thery, mergers and acquisitins JEL Classificatin cdes: C6, D70 We draw n the thery f the cre, a tpic ften taught in cnnectin with general equilibrium thery, t study calitin frmatin using data envelpment analysis (DEA) methds. In the thery f the cre, there is a set {I} f individuals endwed with preferences and initial allcatins f resurces. A grup f {I}, say a subset S I, can imprve upn its members if each member may increase his r her utility by being a member. The cre then cnsists f the allcatin f resurces upn which n calitin can imprve it. Decisin-maing units (DUs) may enter a calitin with ther DUs t imprve their perfrmance (e.g., revenue efficiency r any ther apprpriate perfrmance measure). 2 Deviating frm the standard thery f the cre, ne can frmulate the agents (DUs) by means f DEA r activity analysis mdels. Thus, data prvided, ne can estimate the pssible gains f frming calitins. Having nly a finite number f agents, ne can estimate the best cnditin(s) fr calitin frmatin. Nte that this paper des nt address the allcatin f gains amng participants in a calitin.

Calitin Frmatin and Data Envelpment Analysis 27 First Eample Fr the sae f cnvenience, a simple illustratin will aid in intrducing the tpic f calitin frmatin and DEA. Cnsider three DUs ( =,2,3) each using ne input ( ) t prduce a single utput (y ). The inputs and utputs are hmgeneus, s their sum is well defined, and 3 3, y are ttal input and utput, = = respectively. Table shws hw the three DUs mae up the DEA technlgy. Table DU Inputs and Outputs DU DU 2 DU 3 Output y y 2 y 3 Input 2 3 Intrducing intensity variables, ne fr each DU, z 0 ( =,2,3), allws fr frmulatin f a DEA r activity analysis mdel. In terms f utput sets, ne culd write such a mdel as fllws: 3 3 P ( ) = y: zy y, z, z 0, =, 2,3 = =. One may prve that this mdel has strngly dispsable input and utput (the first tw inequalities) and ehibits cnstant returns t scale, that is P( λ) = λp ( ), λ > 0. The maimal utput that DU ( =, 2, 3) can prduce is estimated as 3 3 F ( ' ) = a zy : z ', z 0, =, 2,3. = = Its efficiency is the rati y' / F( ' ), =, 2, 3. Allwing DU and DU 2 t frm a calitin wuld raise the questin f hw much utput they culd jintly prduce using their cmbined amunt f input, 2 + 2=. The utput wuld be 3 3 3 3 2 2 2 F( 2) = a zy+ zy : z, z 2, + 2 2, z 0, z 0, =, 2,3 = = = =. Frmulating a calitin between DU and DU 2 wuld be beneficial if F( ) > F( ) + F( ). 2 2 Alternatively, ne culd frmulate a weaer cnditin fr calitin frmatin in this case as F( 2) > y + y2, where F( ) and F( 2) are the maimal utput each DU can prduce nt being a member f any calitin, and y and y 2 are the bserved utputs. Similarly, ne culd frm calitins between DUs and 3, between DUs 2 and 3, r amng DUs, 2, and 3. Determining the best calitin wuld invlve cmparing all the alternatives as fllws: F( 2) vs. F( 3) vs. F( 23) vs. F( 23).

28 Calitin Frmatin and Data Envelpment Analysis One culd als mae weaer cmparisns in relatin t bserved utputs as shwn abve. The net sectin invlves generalizing these ideas int multi-utput multi-input technlgies with finite. The General Case Revenue aimizatin N Let inputs R +, utputs y R +, and assume there are =,..., DUs (r firms). The cnstant returns t scale technlgy may be mdeled via DEA r activity analysis as P ( ) = y: zy m ym, m=,...,, z n n, n=,..., N, z 0, =,..., = = where, =,..., are the intensity variables frming the cnve cne f the bservatins (,y ), =,...,. Thin f as the initial endwment belnging t DU, and assume that sme f the inputs may be reallcated amng the DUs, say inputs n =,..., N*. The rest are nn-allcable and stay with their DU. Nte that ne may have N* = N. Althugh here each DU shares the same technlgy, generated by the data (,y ), =,...,, the DUs utput set may differ because they may have different initial endwments, (e.g., n, n ). In the multi-utput frmulatin, ne cannt maimize utputs, s selectin f a methd that allws fr maimizatin is required. Assuming that utput prices ( p ) are nwn, ne may maimize revenue by maimizing m= p y m m subject t a technlgical cnstraint. The maimum revenue fr DU is ' R (, p) = a pm ym : zy m ym, m=,...,, z n n, n=,..., N, z 0, =,..., m= = =. One may estimate the revenue efficiency fr DU as the rati f bserved revenue ' ' ' t maimum revenue R (, p ), that is, ( Ry )/ R (, p. ) m= py m m ' = Ry ' ( ) Net, estimate the revenue efficiency gain DU can mae by frming a calitin with, say, DU 2. Their jint revenue, R(, 2, p ), is estimated as fllws: 2 m m m m m= m= R(, 2, p) a p y + p y = = = zy y, m=,...,, m m z, n=,..., N, * n n = z, n= N+,..., N, * n n z 0, =,...,,

Calitin Frmatin and Data Envelpment Analysis 29 = = z y y, m=,...,, 2 2 m m z, n=,..., N, 2 2 * n n = 2 z, n = N+,..., N, 2 * n 2n z 0, =,...,, + = +, n=,..., N. 2 2 * n n n 2n Perhaps sme further eplanatin is necessary at this pint. Nte the fllwing:. Each DU has its wn intensity variables, z, z, =,...,. 2 2. Bth DUs face the same utput prices; this case can be generalized t 3. One may reallcate the first n =,..., N* inputs t maimize the jint revenue. ' p, =, 2. 2 4. One can cmpare the calitin s revenue, R(, 2, p ), t individual firm revenue, R(,2, p) Ry (, p) + Ry (, p), t determine whether a calitin is beneficial. Evaluating the best calitin ptin fr DU requires a cmparisn with all ther DUs, such as DU 3 thrugh ; fr eample, (, 2, 3), (, 2, 4), and s n. A best calitin eists with being finite althugh it need nt be unique. The General Case Distance Functins aimizatin When data n utput prices are nt available, ne culd add directinal distance functins. The functins are independent f measurement units and, hence, can be aggregated. They als generalize the first eample f adding (scalar) utputs. The apprach is a generalizatin f Jhansen s (972) industry prductin mdel (see Färe & Grsspf, 2004) and can be viewed as an applicatin f benefit thery due t Luenberger (995). First, let P ( ) be an utput set, P( ) = { y : can prduce y }, and g R+, g 0, a directinal vectr. The directinal utput distance functin is defined as fllws: D (, ; ) sup : ( ) ( ) yg = y+ g P { β β }. The directinal distance functin measures the distance, in the directin f frm y t the bundary f the utput set; and is a generalizatin f Shephard s (970) utput distance functin, { θ θ } D (, ) inf : ( / ) ( ) y = y P, where the relatin between the distance functins, fr g = y, is given by DT ( yy, ; ) =. D( y, )

220 Calitin Frmatin and Data Envelpment Analysis One may estimate the directinal utput distance functin using the DEA r activity analysis frmulatin f the utput set P ( ) as D y g β ' ' (,, ) = sup = zy y + β g, m=,...,, m 'm m = z, n=,..., N, n ' n z 0, =,...,. Net, use a distance functin criterin t evaluate the benefits f frming a calitin. Paralleling the revenue maimizatin case, ne may calculate the jint directinal distance functin in the event that DU and DU 2 frm a calitin as fllws: D (, 2; g) = aβ + β = = m m m 2 zy y + β g, m=,...,, z, n=,..., N, * n n = z, n= N+,..., N, * n n z 0, =,...,, = = z y y + β g, m=,...,, 2 2 m m 2 m z, n=,..., N, 2 2 * n n = 2 z, n= N+,..., N, 2 * n 2n z 0, =,...,, + = +, n=,..., N. 2 2 * n n n 2n Emplying the jint distance functin, D (, 2; g), fr the tw individual DU distance functins, (, D ; ) 2 2 y g and D (, ; ) y g, shws whether a calitin between DU and DU2 wuld be beneficial. Again, by evaluating all pssible calitins, ne can find the best gruping amng DUs.

Calitin Frmatin and Data Envelpment Analysis 22 Secnd Eample The secnd eample relates t a case invlving strategic chices f cmpanies. In particular, the prpsed framewr is useful in analyzing hw cmpanies may reach decisins t acquire ptential partners. Because cmpanies eperience increasing difficulty in achieving and sustaining grwth, ften they resrt t frming strategic alliances (e.g., airlines) r acquiring ther cmpanies (e.g., the massive waves f mergers and acquisitins activity in the late 990s). A hypthetical case invlving three bans will aid in investigating the issue further. Assume that the bans use tw inputs, (persnnel) and (capital), t prduce a single utput, (lans and ther investments), as evident in Table 2. Table 2 Ban Input and Output Data Ban A B C y 2 2 2 2 2 The net questin is with which f the ther tw bans, B r C, Ban A shuld frm a partnership. In this case, allw bth inputs t be reallcated. Befre cmmitting t a strategy, Ban A must assess the amunt f redundant resurces that will be a burden shuld it decide t team up with either Ban B r Ban C. The ban culd use surplus resurces t achieve ecnmies f scale r alternatively cut csts by eliminating thse resurces (Dyer, ale, & Singh, 2004). T answer the questin, ne needs t slve tw linear prgramming prblems: LP Prblem : a y + y 2 z + z + z y Ban A 2 3 z 2+ z + z 2 2 3 z + z 2+ z 2 2 3 2 z 0, =, 2,3 z + z + z y Ban B 2 2 2 2 3 2 z 2+ z + z 2 2 2 2 2 3 2 z + z 2+ z 2 2 2 2 2 3 22 z 0, =, 2,3 2 + 3, + 3 2 2 22

222 Calitin Frmatin and Data Envelpment Analysis LP Prblem 2: a y + y 3 z + z + z y Ban A 2 3 z 2+ z + z 2 2 3 z + z 2+ z 2 2 3 2 z 0, =, 2,3 z + z + z y Ban C 3 3 3 2 3 3 z 2+ z + z 2 3 3 3 2 3 3 z + z 2+ z 2 3 3 3 2 3 32 z 0, =, 2,3 3 + 4, + 3 3 2 32 The results shw that Ban A shuld frm a partnership with Ban C, nt with Ban B. In this case, the ttal utput frm Bans A and C is 2.5 units, which is greater than the bserved utput sum f 2 units prduced by any tw ther bans individually r than the maimum jint utput resulting frm a ptential calitin between Bans A and B, which is als equal t 2 units. The requirement in this eample is weaer than in earlier sectins, but it illustrates the pint. Summary The fcus f this paper was t prpse a framewr and present eamples demnstrating hw ne can frmulate and estimate ptimal calitins using DEA methds. At the center f such analyses may be cases invlving strategic chices f cmpanies (e.g., frming alliances r pursuing taevers in the interest f bsting sales revenue and prfits and maimizing sharehlder wealth). Given that crprate histry is fraught with a myriad f failed acquisitins and alliances while taever activity has remained strng as cmpanies eperience even mre difficulty achieving and sustaining grwth, there is strng interest in develping analytical tls t assist cmpanies in maing better deals. The prpsed framewr ffers sme insights int and tls fr helping cmpanies decide whether they shuld acquire ptential partners.

Calitin Frmatin and Data Envelpment Analysis 223 Ftntes In essence, the cre is a generalizatin f the idea f the Paret set. If an allcatin is in the cre, every grup f agents gets sme gain frm trade, and n grup has an incentive t defect (Varian, 992). 2 Eamples f such calitins may include crprate alliances and crprate taevers (Dyer, ale, & Singh, 2004; artynva & Rennebg, 2008; Sudarsanam, 2003). References Dyer, J. H., ale, P., & Singh, H. (2004). When t ally and when t acquire. Harvard Business Review, 82, 08-5. Färe, R., & Grsspf, S. (2004). New directins: Efficiency and prductivity. Bstn, A: luwer. Jhansen, L. (972). Prductin functins. Amsterdam, Netherlands: Nrth-Hlland. Luenberger, D. G. (995). icrecnmic thery. New Yr, NY: cgraw-hill. artynva,., & Rennebg, L. (2008). A century f crprate taevers: What have we learned and where d we stand? Jurnal f Baning and Finance, 32, 248-277. Shephard, R. W. (970). Thery f cst and prductin functins. New Jersey: Princetn University Press. Sudarsanam, S. (2003). Creating value frm mergers and acquisitins: The challenges. Harlw, England: Prentice Hall. Varian, H. R. (992). icrecnmic analysis (3 rd ed.). New Yr, NY: W.W. Nrtn. Authrs Nte Rlf Färe, Department f Agricultural and Resurce Ecnmics and Department f Ecnmics, Oregn State University, Crvallis, OR, USA. Shawna Grsspf, Department f Ecnmics, Oregn State University, Crvallis, OR, USA. Dimitris argaritis, Department f Accunting and Finance, University f Aucland Business Schl, Aucland, New Zealand. Crrespndence cncerning this article shuld be addressed t Dimitris argaritis, Email: d.margaritis@aucland.ac.nz