OraclePeopleSoft Global Competition Review Law Centre Lunch Talk

1
Oracle/PeopleSoftGlobal Competition Review Law
Centre Lunch Talk

• William Bishop, Lexecon Ltd 4th February
  2005

2
Introduction

• Oracle made a hostile bid for PeopleSoft in June
  2003
• Resulted in longest-ever ECMR review (gt12 months)
• Case raises many interesting issues in economics,
  law, politics?
• Lexecon Ltd advised Oracle throughout EC process
  (Lexecon Inc advised PeopleSoft)

3
Timeline

• Oracle announces offer 9 June 2003
• US 2nd request 30 June 2003
• EC notification 14 October 2003
• EC Phase II starts 17 November 2003
• US DOJ challenge 26 February 2004
• EC SO / oral hearing 12 March 1 April 2004
• EC clock stopped for new bid data 14 April 2004
• US trial in San Francisco June - July 2004
• US SF court ruling 9 September 2004
• EC clock restarts 7 October 2004
• EC decision 26 October 2004
• Merger completed 7 January 2005

4
Products and competitors

• Overlap area enterprise application software
  (EAS) particularly human resources (HR) and
  financials (FMS)
• Merging parties are 2 and 3 in EAS worldwide
  SAP is 1 (stronger in Europe than US)
• EAS overall is fragmented more concentrated in
  the high end (largest customers, highest
  functionality products)
• Oracle is also a leader in databases
  complementary to EAS
• Oracle intended to discontinue PeopleSoft
  products eventually

5
The main economic issues

• Product market definition is there a high
  end market? Who competes in it?
• (DOJ/EC disagreed on geographic market EC found
  global market DOJ decided on N. America)
• Unilateral (non-coordinated) competitive
  effects analysis
• Coordinated effects and (briefly)
  vertical/conglomerate issues were also considered
  and rejected

6
Product market definition

• Commissions view High function HR/FMS
  purchased by large customers with complex
  functional needs
• Defined by proxies (1) customers gt10,000
  employees or 1bn revenue (2) deals gt1m net
  license value
• Relied heavily on defining market by reference to
  customer groups nb EAS is sold by individual
  negotiation
• Very difficult to find clear conceptual criteria
  or practical proxies
• Key development in EC case was conclusion on who
  is in the market decision accepted a
  significant fringe (i.e. not 3-to-2)

7
Unilateral effects

• Concern reduction in competition without
  coordination (i.e. merged entity finds it
  rational to raise prices without expecting
  similar reaction from non-merging rivals)
• Typically an issue in differentiated product
  markets, when merging parties are close
  substitutes, and non-merging rivals are less
  close
• Empirical questions how closely do parties
  compete? How closely do rivals constrain them?
• Need to look at bidding data
• SO reached adverse conclusions, but without much
  data

8
Econometric analysis

• PeopleSoft had submitted a very simple (and easy
  to criticise) analysis, claiming discount is
  higher when number of bidders is higher
• Oracle collected detailed information on 600
  bids we carried out extensive econometric
  analysis (50 regressions)
• 2 key questions
• Does number of rival bidders affect Oracle
  discounts?
• Does identity of rival bidders (esp. PeopleSoft)
  affect Oracle discounts?
• We found no systematic evidence of either
  Commission agreed

9
Some thoughts on the economic issues

• A classic gap case (apparently played a role in
  finalising new ECMR)
• Unilateral effects in bid markets what is the
  threshold for a SIEC? And what does theoretical
  merger simulation contribute?
• Market definition for bid markets with
  heterogeneous customers and a continuum of
  characteristics what is a relevant product
  market in this setting? Does it matter?
• Disconnect between geographic and product market
  approaches?
• Loss of choice effect on PeopleSoft installed
  base are they relevant?

10
Why bidding analysis?

• Contracts for EAS software awarded after a
  process of negotiation (i.e. competition takes
  place for the contract)
• Customers normally negotiate with several
  supplies and there are often several rounds of
  negotiation
• Key factor is how does the merger affect
  customers alternatives are the merging
  parties products particularly close and how
  close are the remaining suppliers?
• Commission recognised that sales were made via a
  bidding process

11
Bidding analysis in mergers (1/2)

• Aim is to identity who are the credible bidders
  for a particular type of contract and assess if
  the merging parties are particularly important
  competitive constraints on one another
• The Commission claimed just Oracle/PS/SAP
  credible bidders for high-end HR FMS, and PS an
  important competitive constraint on Oracle
• Market definition typically difficult in bidding
  markets, in particular because of price
  discrimination resulting from negotiation, so
  markets may be defined in terms of customer types
  
• Who competes for different types of contracts can
  be informative, but customers may not invite all
  potential suppliers to bid for contracts

12
Bidding analysis in mergers (2/2)

• Estimate the effects of the merger using the
  variation in the number and identity of bidders
  as a natural experiment
• Merger reduces the number of credible bidders ?
  examine how of bidders affects the discounts
  offered the merging parties and compare
  coefficients on of bidder dummies
• Merger removes particular competitor ? examine
  how the identity of bidders affects the discounts
  offered by the merging parties, and compare
  coefficients on combinations of bidders including
  and excluding the acquired party

13
Bidding analyses done in Oracle/PeopleSoft

• Europe
• SO PeopleSoft regression (Lex Inc)
• September Oracle bidding analysis (Lex Ltd)
• October EU Decision (Commission analysis)
• US trial
• McAfee (DoJ)
• Hausman (Oracle)
• Pre-trial CapAnalysis (Oracle)

14
Lexecon Ltd analyses Response to the SO

• Reviewed PeopleSoft regression analysis. SO
  claimed this analysis showed that the removal of
  one competitor appears to reduce the rebate with
  sic 15 on average
• Problems with PeopleSoft analysis
• Result depends on 8 obs with anomalous margins
  and 3 obs for 3-rival bids which have high
  discounts
• Explained just 8 of variation in discount and
  failed standard diagnostic tests ? excludes
  explanatory variables and possible omitted
  variable bias
• If the 8 anomalous obs excluded and dummies for
  the of rivals included, then only the dummy for
  3 rivals is positive and significant

15
Lexecon Ltd analyses Additional analysis

• Collected new dataset on Oracles sales opps to
  large enterprises
• Econometric analysis
• Effect of of bidders on Oracles discounts, and
• Effect of the identity of bidders on Oracles
  discounts
• Control for characteristics of the sales
  opportunity
• Two measures of the discount used, total and
  discretionary
• Estimate regressions using 11 samples of the data
  proxies for different possible market
  definitions

16
Lexecon Ltd analyses Results

• Data showed that suppliers other than
  Oracle/PS/SAP competed to supply large
  enterprises, but in just a limited number of
  occasions
• No consistent evidence that a higher of bidders
  associated with higher discounts, or discounts
  were higher when there were 3 rather than 2
  bidders
• No consistent evidence that discounts were higher
  when Oracle competed against PS just isolated
  examples which were either due to the presence of
  JDE or reflected just a very small number of
  sales opps.
• Commission obtained same results

17
Issues and problems of any bidding analysis

• Omitted variables common problem with
  cross-sectional data. If the omitted variable is
  correlated with any of the explanatory variables
  then the estimated coefficients are biased and
  inefficient
• Customers may not negotiate with all credible
  suppliers, so does comparing the outcome with and
  without the acquired supplier really capture the
  effect of the merger?