Price behaviour under competition in UK Domestic Electricity Supply

1
Price behaviour under competition in UK Domestic
Electricity Supply

• Monica Giulietti, Jesus Otero and Michael
  Waterson
• Presented by Michael Waterson, University of
  Warwick

2
Basic questions

• How competitive has the domestic electricity
  supply market become?
• What strategies do suppliers appear to be
  pursuing?- firm side of picture
• Related work- GWW on consumer switching behaviour
  in gas, Econ J, October 2005

3
Positive features of the market

• There are now a number of studies focussing on
  pricing, but
• All our prices are transactions prices (no bait
  and switch etc.)
• Nearly everyone buys the product
• They spend a significant proportion of their
  income on it (3)
• We can observe behaviour over a period of time

4
Plan of presentation

• Market opening- institutional features
• Naïve Theory/ Question- competitive?
• Empirical model- descriptive analysis, making
  distinction between search and switching costs
• Empirical results
• Interpretation in the light of theory
• Conclusions

5
(No Transcript)
6
Basic market structure in electricity

• Transmission and distribution regulated,
  generation and supply competitive.
• Since May 1999, all UK consumers have been able
  to choose their electricity supplier.
• Since March 2002 there has been no supply price
  regulation
• Incumbent in each area is a default supplier.
  Other suppliers (entrants) are incumbents from
  other areas, plus the incumbent in gas supply,
  and a small number of independent suppliers.

7
Pricing in supply

• Each firm sells under an (essentially common)
  range of tariff structures, but they have
  different tariffs.
• No special deals to retain consumers- i.e. no
  price discrimination within area allowed.
• For most people (gt2/3, source OFGEM 2004), price
  is the main reason why they would move
  electricity supplier.

8
Switching supplier

• Switching is quite straightforward
• Significant switching has occurred- Nearly 50
  are no longer with incumbent significant churn

9
Questions

• Do prices converge quickly? Do they converge
  at all?- relative prices
• Has the market quickly become competitive?-
  naïve theory position
• What are the implications for search and
  switching costs?
• Assumption- firms know what they doing, in terms
  of pricing strategy.

10
Key magnitudes

• The difference, IM, between the incumbents price
  and the median price offered by entrants.
• The size of the gap, ML, between the lowest
  available price and the median price offered by
  entrants.
• The magnitude of the range, HL, between the
  lowest and highest price offered by market
  entrants.
• The whole difference, IL, between the incumbents
  price and the price set by the lowest price
  supplier.

11
Expectations

• If new switchers face no search costs, we might
  expect ML and HL to be very small if search
  costs fall over time, ML and HL shrink over time,
  and vice versa.
• As N shrinks, HL is likely to decline.
• If there are no switching costs, and the first
  search is very cheap (most people- almost 90
  according to OFGEM- have been contacted by a new
  supplier), we would expect IM to be very small.
• The gap IL may shrink over time.

12
Data

• Essentially, prices
• Bimonthly data for 6 years
• Three bill types (DD, QB, PP)
• Fourteen areas (regional), each with a different
  incumbent
• Two levels of consumption
• Between 18 and 6 companies active, with shakeout
• Significant headroom

13
Difference LM between median and lowest bills
(Direct Debit)
14
Bill range (HL), excluding incumbent (DD)
15
IM and IL, high consumption
16
Descriptive regression model
where Y is ML, HL, IM or IL.
17
Preliminary questions

• Are we dealing with a stationary series?
• Answer essentially, yes in each case.
• All series are stationary around a trend
• There is evidence of some form of structural
  break in Spring 2002- we split the sample at this
  point
• Final price controls removed, plus rise in fuel
  input prices

18
Restricted model. Trend interaction by region and
product. Jun. 02Dec. 06 (extract)
Regressors ?Y ?ML ?Y ?ML ?Y ?ML ?Y ?ML ?Y ?HL ?Y ?HL ?Y ?HL ?Y ?HL
Feb. 99Apr. 02 Feb. 99Apr. 02 Jun. 02Dec. 06 Jun. 02Dec. 06 Feb. 99Apr. 02 Feb. 99Apr. 02 Jun. 02Dec. 06 Jun. 02Dec. 06
Coeff. t-Stat Coeff. t-Stat Coeff. t-Stat Coeff. t-Stat
Y(-1) -0.33 -8.63 -0.52 -12.30 -0.18 -6.29 -0.25 -7.06
NFIRMS 0.10 1.69 0.66 7.06 0.35 3.19 1.19 8.05
Group 1
TrEADD -0.13 -4.76 0.08 3.61 -0.19 -3.12 0.34 8.16
TrEMDD -0.09 -2.30 0.08 3.45 -0.17 -2.28 0.31 7.33
19
Restricted model. Trend interaction by region and
product. Jun. 02Dec. 06 (cont)
Regressors DY?DIM DY?DIM DY?DIM DY?DIM DYDIL DYDIL DYDIL DYDIL
  Feb. 99Apr. 02 Feb. 99Apr. 02 Jun. 02Dec. 06 Jun. 02Dec. 06 Feb. 99Apr. 02 Feb. 99Apr. 02 Jun. 02Dec. 06 Jun. 02Dec. 06
  Coeff t-Stat Coeff. t-Stat Coeff. t-Stat Coeff. t-Stat
Y(-1) -0.2 -3.69 -0.31 -9.94 -0.47 -9.96 -0.3 -10.34
NFIRMS 0.09 2.55 -0.27 -2.38 0.14 2.35 0.54 3.69
Group 1                
TrEADD 0.15 3.44 -0.01 -0.2 0.22 5.89 0.11 3.11
TrEMDD 0.13 3.35 0 -0.06 0.23 4.85 0.11 3.18
20
Results (DD case)

• Results differ across the periods up to April
  2002 and June 2002 to December 2006.
• ML and HL decline in the first period. However,
  they rise, particularly HL, in the second period.
• IM rises slightly in the first period and remains
  constant in the second.
• As N falls, HL and ML decline. However, IM rises
  as N falls in second period.
• Speed of adjustment- all variables are trend
  stationary. Rate of convergence speeds up in
  second period.

21
Explaining the results

• Naïve (Bertrand- type) theory does not work
• Incumbency confers an advantage, but other
  factors in addition. Specific prediction on IL
  shrinking not borne out
• Some unexpected results, over time
• Search models
• Anderson de Palma, passive search
• Stahl-type mixed strategy modelling

22
Passive search model?

• Anderson- de Palma model is in some ways
  promising Consumers have little or no prior
  experience of the product and consumers search
  passively. Positive relationship between HL and
  N, also.
• But- firms are both good and bad buys over
  time
• (Lach) No significant correlation between Feb 99
  price and t price after around 12 months.

23
Randomising of prices?- reasonably so
24
Mixed strategy equilibrium model

• Mixed price equilibrium search models (Stahl et
  al) imply in these circumstances that
• Price dispersion persists
• Randomising of prices
• Number of suppliers may or may not influence
  dispersion
• If proportion doing complete search rises,
  average price falls.

25
Mixed strategy equilibrium reconsidered

• However, average prices do not show a particular
  tendency to fall (n.b recent experience)
• But internet usage has increased significantly
  over time, and proportion searching for energy
  prices through this method has also increased
• So a puzzle?

26
What about reswitching?

• Recall that only a minority of switches is from
  the incumbent to an entrant
• Hence, the decision making process for an entrant
  becomes more complex

27
Modelling an entrant after the initial period
28
Lognormal(1,4) simulations n15 and n6

• Note that as time passes, industry moves from l1
  to l0.5 n falls from around 15 to 6.
• Implication is that average price and range
  change over time
• two separate forces.

29
Concluding remarks

• Despite search costs falling over time and people
  used to switching, plus very active switching
• Significant dispersion in prices across entrants
  and between entrants and incumbent persists
• Firms can take advantage of their captured as
  well as captive customers (BG)
• We do not necessarily expect further price
  convergence, due to mix of forces.
• Is the market competitive??