CCRP Summer Workshop 2008

JULY 11, 2008

City University’s Centre for Competition and Regulatory Policy summer workshop took place today and yesterday and I was there to present The Control of Porting in Platform Markets. As well as presenting I had the chance to take some ‘impressionistic’ notes on some of talks which are included below.


Session 1: Telecoms and Postal Services

PAUL SMITH - CEPA: Defining the universal postal service

CARLO REGGIANI – UNIVERSITY OF YORK: Network neutrality and non-discriminatory issues: An economic analysis

  • 2 recent papers (2007): Economides and Yal + Yermelo and Katz
  • 2 sided-model
  • n-firms providing platform (telecoms)
  • network externalities both sides
  • Questions:
    • do telcos set prices on both side
    • What is form of the competition
    • net neutrality is always bad so why used

Session 2: Competition issues

RUFUS POLLOCK - CAMBRIDGE UNIVERSITY: The control of porting in two-sided markets

DAVID GILL – UNIVERSITY OF SOUTHAMPTON (with John Thanassoulis): The impact of bargaining on markets with price takers: Too many bargainers spoil the broth

  • What happens if some consumers bargain a discount from list prices

  • Some proportion of consumers z do not bargain

    • exogenous but endogenized later on
    • Cournot competition for these guys (with Bertrand this all goes wrong …)
  • Of those that do bargain some get one quote some get multiple (Bertrand from multiple)

    • trade-off getting monopoly from single quote guys vs. purchase from multi-quoters
    • From Judd + Burnett 1983
    • [ed: is there a cost for getting quotes]
    • [ed: Very like Baye and Morgan and resulting in similar mixing results)
  • Firms anticipate that higher list prices raises profits from bargainers

  • So as number of bargainers go up firms raise list prices

  • Results

    • As bargainers prop. increase price-takers do worse
      • Waterbed effect + fact that
    • Existing bargainers CS decreases as prop. bargainers rises
    • Swapping consumers (price-taking to bargaining) benefit
    • Overall effect: ambiguous
      • Overall negative and most bargainers get only one quote
      • Overall positive if most bargainers get multiple quotes
  • Then endogenize number of bargainers by assuming some intermediate types who face cost c of bargaining

    • Results similar
    • Still do not endogenize choice of number of quotes – discussed in paper but not done
  • Comments:

    • Waterbed effect: what if firm entry (i.e. zero profit condition) then better prices for bargainers => worse prices for price-takers
    • Baye, Morgan

GERT BRUNEKREEFT – JACOBS UNIVERSITY BREMEN: Ownership unbundling of the German electricity TSOs – A social cost benefit analysis

VINCENT RIOUS – SUPELEC (with Jean-Michel Glachant, Yannick Perez and Philippe Dessante): The diversity of design of TSOs

STEPHEN WOODHOUSE - POYRY: Wind generation – no limits?

  • Everyone is signing up to incredibly optimistic renewable and CO2 targets.

  • For UK wind is essential as we have a lot of it compared to any other renewable options

  • However wind has major delivery issues and conventional wisdom is that its max penetration is 10%

  • Problem is:

    • wind can be irregular
    • (more significant) demand shows pronounced fluctuations over the day while renewables don’t (on average). This means that your backup capacity to deal with peak load make renewables on avg. v. expensive.
    • why this debate about whether feasible or not – why can’t we simply price carbon efficiently
    • like a man who has a dislocated shoulder and spends all his time trying to fix the pain this causes in his hip rather than sorting out his shoulder

Session 4: Evaluation of Regulation and Competition Institutions

GORDON HUGHES – UNIVERSITY OF EDINBURGH: Efficiency frontiers, stranded assets and the X-factor for telecoms network operators

  • Setting the X in RPI - X
  • Stochastic frontier analysis
  • Look at 68 US local exchange carriers (data from FCC)
    • Current costs from historic accounts
    • Stranded assets (from switch to digital)
  • Structural break in 2000
    • To 1999 costs falling at -3.3%. From 2000 falling at -2.1%
  • Stranded assets affect costs: cumulative impact of 5% annual decline in switched line equivalent to a cost increase of ~2.6% per year
  • Slow convergence towards frontier: ~1.3% per year
  • Accounting vs. economic cost important
    • Accounting cost: ~ -1.7% per year (post 2000)
    • Economic cost: ~ 1.7% per year (post 2000)
  • RPI-X:
    • using accounting costs: X ~ 1.5-2.5%
    • using economic costs: X ~ -0.5 - 2.0% (i.e. -ve and prices rise faster than inflation)
    • In europe can justify + ~2.5% to X but this will all over time.

JOHN CUBBIN - CITY UNIVERSITY (with Jon Stern, Federica Maiorano and William Gboney): What can we learn from economic studies of infrastructure regulatory policies?


Session 1: Transport

ANNE YVRANDE-BILLON - UNIVERSITY PARIS SORBONNE (with Miguel Amaral and Stephane Saussier): Does competition for the field improve cost efficiency? Evidence form the London bus tendering model

  • Competition for market
  • Idea is that competition raises bids (whether charges for providing service or payment for right to run it)
  • Little empirical testing
  • Several confounding factors
    • Winner’s curse: can happen in common-value and in private value auctions if bidders systematically under-estimate their own costs (i.e. over-estimate their own values)
  • Renegotiation effect: a bid not be allowed if not good enough even if it wins (implies more aggressive bidding)
  • Entry effect: Larger number of expected bidders might discourage entry.
  • Existing papers on impact of no. of bidders on outcome
    • Branman et al (1987), Thiel (1988), Dalen + Gomez-Lobo (2001), Hong + Shum (2002) – find strong winner’s curse, Nunez + Athias (2006)
    • Do not control for other extra factors
  • France vs. London (Amaral, Saussier + Yvrande 2008)
    • French Urban Public Transport sector
    • declining productivity, huge deficit – basically a disaster
    • tendering model (for buses):
      • No clear selection criterion (intuitu personae) – right enshrined in law by vague definition of the ‘collective welfare’
      • No regulator
      • Few bidders (av 1.4)
      • 66% of auctions with only one bidder
      • Incumbent advantage (~88% renewed)
      • Collusion (fined by Comp. Commission 2005)
    • Bus auctions in France are for complete networks while for UK they are for routes
    • This excludes Paris as Paris directly administered
  • UK model
    • Bus operation auctioned on route-by-route basis
    • Bids are annual price for service provision
      • Revenues occur to authority – so service provider has no demand risk (just ‘industrial’ risk)
    • Selection criterion ‘best economic value’ but:
      • Qualitative factors count (e.g. reputation, quality)
      • Discretionary power of the regulator (TfL) – may not select the lowest bidder if a) do not think firm can deliver b) would result in more than 20% market share c) …
      • A public benchmark exists (what was the old public operator)
    • Auction format: combinatorial first price auction. Aims to:
      • Encourage participation of small operators by unbundling the network
      • Benefit from coordination and scale and scope via package format
  • Regarding initial concerns:
    • These are private value auctions so less risk of winner’s curse
      • Cantillon + Pesendorfer (2006): private information about opportunity costs
      • [ed: not sure here. would seem likely that there is a strong common component here]
    • No renegotiation of contracts: short term contracts and strong regulator
  • Dataset: all auctions between March 2003 and May 2006 (294 individual routes)
    • all on the regulator’s website!
    • other information about the transport network
  • Summary info:
    • Constant over time (unlike France)
    • Around avg 3 bidders per auction
    • Only 20% of auctions have one bidder
  • Basic regression:
    • Av cost per mile (cpm) does decline with number of bidders
    • But clear endogeneity problem as av. bus miles correlated with number of bidders and costs
    • Deal with this by using predicted number of bidders based on number of operators in the vicinity of the route in the previous period.
    • However correlating actual and predicted number of bidders find -ve correlation (suggests people enter (and bid high) when the number of expected bidders is low and vice-versa)
      • confirms endogeneity of entry
  • Results:
    • N effects bids in the way we would expect
    • Competition effect larger than (deterred-)entry effect
  • Discussant comments:
    • Still carry some demand risk because demand may impact on cost of operation
    • Data on congestion would be useful

ALBERTO GAGGERO- UNIVERSITY OF ESSEX (with Claudio Piga): Pricing and competition on the UK- Irish aviation market

  • Background
    • UK-Irish aviation market is largely dominated by Aer-Lingus (EI) and Ryanair (FR)
    • Ryanair launched takeover in 2006 but was blocked in 2007
  • Test whether there is impact of competition
    • Plus a study with European data (most from US)
  • Data
    • 84k flights
    • EI: 30%, FR: 55%, next biggest 10%
    • ~ 25 routes
    • Using web spider have full fares dataset
    • CAA: available seats, sold seats, flight frequency (aggregated)
    • Distance in km between 2 endpoints
  • Put in most variables you could think of
  • Endogeneity issues:
    • pricing and market structure may be simultaneously determined so do IV
    • IV approaches mostly based on the fact that the more likely one serves both ends of a route the more likely one serves that route
  • Results:
    • (Surprisingly) market shares variables go wrong way (higher market share lower prices)
    • This holds with IVs or without
    • Different IVs do affect size of negativity but do not change the sign
    • (Route) market share up 1% reduces fares by 0.19% (Borenstein IVs) or 0.5% (their own IVs)
    • Check robustness (e.g. pooling all London airports)
    • All other regressors economically and statistically significant and of right sign

Session 2: Finance

ENRIQUE BENITO - FSA: Size, growth and bank dynamics

  • Background
    • Banking in Europe has changed a lot (lots of deregulation)
    • Size in banking is important
    • Little examination of size of banks
    • General increase in concentration (more big banks)
    • Data on Spanish banks 1970-2006
  • Traditional literature:
    • X-section regression to explain current sizes as function of underlying factors (and hence trends over time in size and concentration driven by these underlying factors)
  • Here focus on classic Gibratian stochastic growth process (LPE)
    • S(i,t) = S(i,t)^beta exp(mu(i,t))
    • mu(i,t) = N(alpha(i) + delta(t), sigma)
  • Predictions from LPE
    • P1: beta = 1
    • P2: No persistence in growth across periods (no correlation across periods)
    • P3: Variability of growth rates is independent of size
  • If these hold (strong all 3, weak just P1) then growth rates of banks follow random walk with drift
  • Data
    • Annual data for all banks
    • Reliable data maybe from 1980 so do everything both 1970-2006 and 1980-2006
    • Include firms that exit plus mergers [ed: not quite sure how they deal with mergers exactly]
  • Results:
    • Beta less than 1 (significantly but not by much). Some (IMO) weak evidence that is has increased a little bit in more recent periods
    • Rho (measure of convergence) is significantly above 1 (which implies previous periods growth predicts growth today)
    • Heteroscedascity: yes (size matters)
    • Variability of growth: larger banks have more stable growth
    • So reject LPE over whole period but may be converging towards it over time
  • Conclusion:
    • Size-growth relationships change over time
    • Converging towards LPE => more skewed size distribution in future (more concentration)

KAI KOHLBERGER - FSA (with Richard Johnson): Has MCOB regulation affected the suitability of subprime mortgage sales?

  • Did introduction of new regulations (MCOB) affect mis-selling
    • Mortgages should be suitable (explicit defn)
  • Approach
    • Look at arrears rate 12 months after sale
  • Data
    • 15 firms, 590k mortgages
    • Regressions with 300k observations (due to missing values – check this is not systematic)
    • FSA Product sales database (PSD)
    • Macro vars
    • Subprime defined as in PSD
  • Find no impact on arrears rates discernible from policy change