EE Dresden University of Technology Chair of Energy Economics and - - PowerPoint PPT Presentation

• 1 -

EE²

Evaluating the Efficiency Effects of Industry Consolidation

Evidence from US Interstate Pipeline Companies

Borge Hess

INFRADAY 6 October, 2007, Berlin

EE²

Dresden University of Technology Chair of Energy Economics and Public Sector Management

• 2 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 3 -

EE²

Introduction

There are different business strategies that prevail in the US natural gas interstate pipeline industry that become increasingly interesting

• 1. Increasing number of acquisitions per years
• 2. Formation of big holding companies
• 3. Cooperation investment in pipeline (Joint Ventures)

Are those business strategies successful?

• 4 -

EE²

The US Natural Gas Industry

Source: El Paso (2006): GHG Inventory Development for Natural Gas Pipeline Company

• 5 -

EE²

US and Canadian Natural Gas Pipelines

Source: CEPA (2004): Presentation to Minister

• f Environment Canada

• 6 -

Why do Firms merge?

Diamond and Edwards (1997) emphasized five major causes:

• 1. Economic efficiency in form of cost savings by synergy effects;
• 2. Defensive motives
• 3. Diversification
• 4. Growth and personal aggrandizement
• 5. Market power

+ Supply security (gas fired electricity generation, natural gas supplier)

Efficiency: production function, define the relationship between the inputs

and outputs. Represents the maximum output attainable from each input level, reflects current state of technology; firms operating on the frontier technically efficient.

• 7 -

EE²

State of the Literature

Efficiency estimation of natural as transmission companies

• Sickles, Streitwieser (1992):
• 14 US interstate Gas Transmission Companies (1977-1985), SFA, DEA, Production function
• Findings suggested the introduction of the Natural Gas Policy Act of 1978 to affect a decline in

technical efficiency

• Granderson, Linvell (1999):
• 20 US interstate Gas TSOs (1977-1987), SFA, DEA, Cost function
• Quite similar ranking of firms of DEA and SFA efficiency scores

Related work on mergers only concerning electricity sector:

• Nillesen, Pollitt and Keats (2001) and Nillesen and Pollitt (2001)
• Kwoka and Pollitt (2005 and 2007): DEA and Tobit regression on panel data set (78

distributors; 1994-2001) buying firms are winners / targets are losers of a merger

We use parametric Stochastic Frontier Analyses (SFA) to analyze the effect

• f business strategies (mergers, holding, Joint venture) on technical

efficiency

• 8 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 9 -

EE²

Stochastic Frontier Analysis (SFA)

X Input Output Y

XA noise effect vi <0 noise Effect Vi>0 XB YB YA

Estimation by using ML estimation 1) Obs. ( ) are controlled for random noise 2) Difference of costs ( ) to minimum cost function is inefficiency

YSFA = βX ± vi - ui

vi~iidN(0,σv

2) ui~iidN+(µ,σu 2)

Deviations are either due to noise (vi) or due to inefficiency (ui)

inefficiency effect ui

Maximum Production Function

• 10 -

EE²

Model Specification

Carrington, Coelli and Groom (2002) discussed physical vs. monetary data models in form of capital measures in the gas industry

• Physical (Pipeline Length):

+ Easily to get

• Cannot capture the total capital equipment
• Difficult to account for differences, e.g. age, quality and

composition (sizes or materials used)

• Monetary measures (Transmission Assets):

+ Account for the total equipment

• Difficulties with different accounting standards

Discussion can also be related to the correct output measue (gas delivered

• vs. total revenues)

Companies in the sample use similar accounting methods/standards We specify Monetary data models due to their advantages

• 11 -

EE²

Models Used

X Holdings: different companies X Merger dummies: Time periods X X X X X X X Model 2 X X X X X X X X Model 3 X X X X X X X X X Model 4 X Time trend X Offshore pipeline X

• Compr. Station’s

intensity STRUCTURE X Total Revenues OUTPUT X Time trend Joint Venture Holding dummy X Merger dummies: Time path STRATEGIES X Transmission Assets X OPEX INPUTS Model 1

• 12 -

EE²

Functional Form

it it t i OFFSHORE it INTENSITY CS it ASSETS it OPEX it

v u TIME OFFSHORE INTENSITY CS ASSETS OPEX REVENUES − + + + + + + = β β β β β β _ ln ln ln

_

∑

+ + =

m m m t it

it

d t δ δ δ µ

Applying SFA on Cobb-Douglas production function within a TE Effects Model (Battese/Coelli 1995) A Firms’ Inefficiency is explained in a simultaneous step

• 13 -

EE²

Data

Data come from US federal energy regulator FERC – Form 2/2a data

• 47 interstate natural gas pipelines over 10 years (1996-2005)
• Balanced panel with 470 obs.
• Heterogeneous sample but covers ca 86% of interstate pipeline network and 93% of pipeline

capacity in 2005

• 46 mergers and 13 holding companies are analyzed
• Holdings companies incorporated cover cover about 65% and 70% of total pipeline network and

capacity, respectively

• FERC is accounting data for each pipeline operator separately whether merged or not

73,400 3.01 0.08 1,220 4,077 200 1,130

• Std. Dev.

2,163 0.073 0.00 10,7 25 0.06 759 Min 393,000 14.9 0.43 6,000 16,666 907 5,950 Max 60,700 OPEX (tsd $) 3.01 Peak Delivery (Mio. Dth per day) 0.21 Compressor Station’s Share of Total Transmission Assets 1,180 Total Transmission Assets (mio. $) 3,905 Pipeline Length (Miles) 203 Total Revenues (mio. $) 949 Total Deliveries (Mio. Dth = 1bn cf)) Mean Explanation

• 14 -

EE²

Timing of Mergers and Cooperative Structure

Timing of Merger

2 4 6 8 10 12 14 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Data come from SEC (Securities and Exchange Commission) and various firms’ websites 46 mergers and 13 holding companies are analyzed Holdings companies incorporated cover about 65% and 70% of total pipeline network and capacity, respectively 70% of all observations are related to holding structures

Sample Coverage (in %)

5 10 15 20 E l P a s

• W

i l l i a m s K i n d e r

• M
• r

g a n S p e c t r a / D u k e Q u e s t a r C

• a

s t a l E n r

• n

O n e

• k

C e n t e r p

• i

n t C h e v r

• n

S

• u

t h e r n U n i

• n

T r a n s c a n a d a N i S

• u

r c e J

• i

n t V e n t u r e

• 15 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 16 -

EE²

Estimation of the Production Function

Inputs are significant and have the correct sign All models show similar results

• Assets have highest revenue

elasticity, as expected

• The higher the share of

compressor station assets on total assets, the higher is the revenue

• Revenue reduction by 3-4%

each year

• Offshore pipelines have

significantly lower revenues

• Might be due to small

distance pipelines

well specified production function

• 78.37
• 48.19
• 315.58
• 73.12

Log Likelihood

Significance 1%-, 5%-, 10%-level: ***,**,*; SE in parentheses.

3.33*** (1.11) 0.43*** (0.06) 3.33*** (1.40) 2.46*** (0.84) σ

2

• 0.04***

(0.00)

• 0.03***

(0.01)

• 0.04***

(0.00)

• 0.03***

(0.00) TIME

• 0.73***

(0.05)

• 0.80***

(0.05)

• 0.65***

(0.05)

• 0.63***

(0.05) OFFSHORE 0.43** (0.19) 0.36* (0.20) 0.59** (0.20) 0.61*** (0.18) CS_intensity 0.82*** (0.02) 0.81*** (0.02) 0.82*** (0.02) 0.82*** (0.02) lnASSETS 0.12*** (0.02) 0.12*** (0.02) 0.13*** (0.02) 0.13*** (0.02) lnOPEX 0.12 (0.24) 0.38 (0.26) 0.04 (0.25) 0.10 (0.21) constant Model 4 Model 3 Model 2 Model 1 Coefficient

• 17 -

EE²

Results from Merger Analysis

POST-MERGER PRE-MERGER 3.92*** (1.50) post-merger 5.18** (2.05) pre-merger

• 0.05 (0.03)

TIME

• 14.43** (6.23)

Constant Model 2 Model 2 4.13*** (1.34) 9 years after 2.68*** (0.82) 1 year before 1.69* (0.95) 8 years after 3.06*** (0.99) 2 years before 1.74* (0.96) 7 years after 3.72*** (1.13) 3 years before 3.65*** (1.23) 6 years after 2.77*** (0.86) 4 years before 3.15*** (1.02) 5 years after 1.89** (0.76) 5 years before 3.92*** (1.29) 4 years after 4.82*** (1.50) 6 years before 2.75*** (0.94) 3 years after

• 0.77 (0.58)

7 years before 2.24*** (0.73) 2 year after

• 2.45* (1.43)

8 years before 2.58*** (0.85) 1 year after 4.75*** (1.61) 9 years before

• 0.13* (0.07)

TIME

• 10.27***(3.60)

Constant Model 1 Model 1

Inefficiency is decreasing

• ver time remarkably

Model 1 shows almost always significant positive values fro the time path dummies

• Overall effect cannot

be evaluated Model 2 shows decreasing but still positive values Acquired pipelines are less efficient than non-acquired firms, but after acquisition the effect reduced

• 18 -

EE²

Results from Cooperation Analysis

• 0.02 (0.03)

TIME 7.79*** (2.36) Holding

• 0.23*** (0.08)

TIME 0.88*** (0.28) Joint Venture

• 12.90***(4.38)

Constant Model 4 Model 4

• 2.07** (1.00)

Enron 0.13 (0.19) NiSource 0.09 (0.36) Coastal

• 0.37 (0.87)

Transcanada 1.24*** (0.21) Questar 0.43 (0.51)

• South. Union
• 0.60 (0.77)

Spectra/Duke

• 3.96*** (1.31)

Chevron 0.59*** (0.16) Kinder-Morgan 0.75*** (0.27) Centerpoint

• 3.12*** (0.98)

Williams

• 0.11 (0.63)

Oneok

• 0.34* (0.21)

El Paso 0.68*** (0.22) Joint Venture

• 1.04*** (0.30)

Constant Model 3 Model 3

Inefficiency is decreasing

• ver time remarkably

A Joint Venture appears to rive down the efficieny (e.g. multiple interests) Model 3 shows heterogeneous picture of inefficiency

• f holding companies
• Williams and Chevron

are very efficient (due to Oil experience?) Model 4 shows average evidence for a large efficiency drop by being part f a holding/parent company

• 19 -

EE²

Efficiency Estimates

Sample of 470 obs. Very similar efficiency results across all models Low variance but also some very bad performer Average technical efficiency of about 80% High correlation of over 90%

0.97 0.94 0.04 0.86 0.80 Model 2 0.96 0.93 0.04 0.83 0.78 Model 3 0.96 0.97 95th Percentile 0.93 0.93 Maximum 0.04 0.04 Minimum 0.85 0.84 Median 0.79 0.78 Mean Model 4 Model 1 Statistics

• 20 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 21 -

EE²

Conclusion

Presenting a fresh approach for analyzing business strategies with respect to technical efficiency

• Estimate technical efficiency of interstate natural gas pipeline companies from the US for

1996-2005

• Applied a robust one-stage SFA (Battese, Coelli 1995) with a Cobb-Douglas production

function

Pipeline acquisitions lead to an increase in efficiency but non-merging firms still perform better Joint ventures have lower efficiency than pipelines fully owned by one company Holding structures on average lead to lower efficiency, but firms with experience in the oil pipeline industry perform better We cannot find evidence for successful business strategies of acquisitions, joint ventures or holding structures Further work:

• Controlling for unobserved heterogeneity, scope effects gas&power and gas&oil

• 22 -

EE²

Thank you for your attention. Comments and questions are welcome.

Contact: Borge Hess borge.hess@tu-dresden.de

EE²

Dresden University of Technology Chair of Energy Economics and Public Sector Management

• 23 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 24 -

EE²

Literature (selected)

Battese, G. E. and Coelli, T. J. (1995), A model for technical inefficiency effects in a stochastic frontier production function for panel data, Empirical Economics 20, 325-332. Berechman, J. (1993), Public Transit Economics and Deregulation Policy, Studies in Regional Science and Urban Economics, North-Holland, Amsterdam. Coelli, T. J. (1996) A Guide to FRONTIER Version 4.1: A Computer Program for Stochastic Frontier Production and Cost Function Estimation, CEPA Working Paper 96/7, Department of Econometrics, University of New England, Armidale NSW Australia. Coelli, T. J., Prasada Rao, D.S., O’Donnell, C., and Battese, G. E. (2005), An introduction to efficiency and productivity analysis, Second Edition, Springer, New York. Diamond, J. and Edwards, J. D. (1997), Mergers, Acquisitions and Market Power in the Electric Power Industry, California Energy, Commission, Research Paper, EIA (1999), The changing structure of the electric power industry 1999: Mergers and Other Corporate Combinations, EIA (2000), The changing structure of the electric power industry 2000: An Update, Herrero, I. and Pascoe, S. (2002), Estimation of technical efficiency: a review of some of the stochastic frontier and DEA software, Computers in Higher Education Economics Review, 15 (1) Kumbhakar, S. C., Ghosh, S. and McGukin, J. (1991), A Generalised Production Frontier Approach for Estimating Determinants of Inefficiency in US Dairy Farms, Journal of Business and Economic Statistics, 9: 279-286. Kumbhakar, S. C. and Lovell, C.A. K. L. (2000), Stochastic Frontier Analysis, Cambridge University Press, Cambridge. Kwoka, J., Pollitt, M. (2005), Restructuring and Efficiency in the U.S. Electric Power Sector. Kwoka, J., Pollitt, M. (2007), Restructuring and Efficiency in the U.S. Electric Power Sector. Northeastern University Working paper 07-001. Lowry, M.N., Getachew, L. and Hovde, D. (2005), Econometric Benchmarking of Cost Performance: The Case of US Power Distributors, The Energy Journal, 26(3), 75-92 Pitt, M., and L. Lee (1981), The Measurement and Sources of Technical Inefficiency in Indonesian Weaving Industry, Journal of Development Economics, 9, 43-64. Ray, D., Stevenson, R., Schiffman, R., and Thompson, H. (1992), Electric Utility Mergers and Regulatory Policy, National Regulatory Research Institute, Occasional Paper 16, 27-41. Reifschneider, D. and Stevenson, R. (1991), Systematic Departures from the Frontier: A Framework for the Analysis of Firm Inefficiency, International Economic Review, 32, 715-723. SEC (2006), SEC website, http://www.sec.gov/divisions/investment/opur/regpucacompanies.htm Simar, L. and Wilson, P. W. (2003), Estimation and inference in two-stage, semi-parametric models of production processes, Discussion Paper (0307) of Institute of Statistics, University Catholique de Louvain. (forthcoming in Journal of Econometrics). Technical Inefficiency and Productive Decline in the Sickels, Streitwasser (1992): U.S. Interstate Natural Gas Pipeline Industry Under the Natural Gas Policy Act

• 25 -

EE²

Agenda

• 1. Introduction
• 2. Methods
• 3. Empirical Results
• 4. Conclusion

Literature Appendix

• 26 -

EE²

Is the Efficiency Shift due to Economies of Scale?

1 1

ln ) , ( ln

− =

⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ∂ ∂ = ∑

M m m c

q q w c ε

Merging parties have higher ES than the average/non- merging parties Merging parties increase their SE after merging But: Buyer have higher increase in scale efficiency than the target Efficiency gains might be due to scale effects Efficiency losses have its sources in technical and/or allocative efficiency Measuring economies of scale by the inverse

• f the sum of cost elasticities of outputs:

ES do (not) exist, if ε > (<) 1 Measuring economies of scale by the inverse

• f the sum of cost elasticities of outputs:

ES do (not) exist, if ε > (<) 1

1 1

ln ) , ( ln

− =

⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ∂ ∂ = ∑

M m m c

y w y c ε

Economies of Scale

1.03 1.035 1.04 1.045 1.05 1.055 1.06 1.065 Model 1 Model 2 Model 1 Model 2 Model Specification Economies

Target Buyer

• 27 -

EE²

Results: Explaining Efficiency

• 28 -

EE²

Parametric Approach of Efficiency Analysis

Applying Stochastic Frontier Analysis (SFA)

• Use of one-stage procedure to estimate inefficiency and its sources simultaneously

(Technical Efficiency Effects Model by Battese/Coelli, 1995)

• Random Effects Model á la Pitt and Lee (1981) --> heterogeneity is treated as inefficiency
• Using translog cost function for a panel data set with mean correction
• 2 outputs: electricity delivered, customer numbers
• 2 inputs prices: cost of capital and labor
• Network density (customer number per unit of assets)
• Distributors which are operating in densely settled area have cost advantages
• Software FRONTIER 4.1

Accounting for mergers

• Three groups of firms: buying firms, acquired firms, non-merging firms
• Dummies for timing of mergers

• 29 -

EE²

Results: Cost function

• 30 -

EE²

Results: Explaining Efficiency I

• 31 -

EE²

Results: Explaining Efficiency II

• 32 -

EE²

Components of Efficiency

Source: Stefanou (2006): Lecture notes, Wageningen Summer School

Technical Inefficiency Allocative Inefficiency

• 33 -

EE²

Parametric Methods

(Pollitt (2001), 6)

n nx

x C β β β + + + = K

1 1

ˆ

Output X Costs C A B F E D OLS: COLS = βx + βOLS ± vi SFA: CSFA = βx + βSFA ± vi + ui COLS: CCOLS= βx + (βOLS -vmax) + ui

C Costs X Output, Input price V stochastic error (White noise) U Inefficiency Efficiency of firm BSFA = EF/BF

Parametric: functional form Easiest form linear:

) , , ( ˆ

1 n

x x f C K =

COLS SFA OLS