Post on 28-Mar-2015
Intra- and Inter-Format Competition in Grocery Retailing
Catherine Ball
ESRC Centre for Competition Policy
University of East Anglia
20/02/09
Introduction– Research questions– Literature
Econometric model– Methodology– Data
Results Conclusions Further work
Outline of Presentation
Research Questions/Literature
How effective are small supermarkets and specialist stores at providing a competitive constraint to large supermarkets?
– Several CC/OFT studies make the assumption that ‘larger’ supermarkets constrain ‘smaller’ supermarkets but not vice versa.
• Cleeren et al (2008) show that discounters can constrain supermarkets in Germany.
Which type of supermarkets (small, urban or large,out-of-town) are the most ‘harmful’ to the existence of specialist stores?
– Smaller, town-centre supermarkets might be less ‘harmful’ to specialist stores – increase footfall.
Is it useful to measure variation in product range, both in terms of range of product categories and range within categories, in a two dimensional product space?
– Increased scope for product differentiation decreases competitive pressure in a market Seim (2006) – could this extend to differentiation in product range carried?
Market Summary
Large supermarkets– 4 major players
• Tesco (supermarket and Extra formats) (27.6% all formats)• ASDA (14.1%)• Sainsbury (supermarket format) (13.8% all formats)• Morrisons (9.9%)
Market Summary
Large supermarkets– 4 major players
• Tesco (44,000)• ASDA (65,000)• Sainsbury (26,000)• Morrisons
Market Summary
Smaller supermarkets/convenience stores– Main players
• CGL/Somerfield (3.8 and 3.9% respectively 2007)• Waitrose (3.3%)• M&S (3.8%)• Tesco (Express and Metro formats) (27.6% all formats)• Sainsbury (Local and Central formats) (13.8% all formats)• Budgens• Iceland (1.5%)• Aldi/Lidl (2.8% combined)• Others (15.5%)
Market Summary
Smaller supermarkets/convenience stores– Main players
• CGL/Somerfield (15,201 – CGL only)• Waitrose (24,500)• M&S (5,000 – food only)• Tesco (Express and Metro formats) • Sainsbury (Local and Central formats) • Budgens (6,600)• Iceland (3,420)• Aldi/Lidl• Others
Market Summary
Specialist stores– Main types:
• Fishmongers • Greengrocers• Butchers• Bakers• Off-licences
Multiproduct Firms
Most retailers are multiproduct firms:
Two issues-– Internalisation of agglomeration effects (Shaked and
Sutton, 1990)• Expansion effect• Lower for specialist stores
– Product differentiation in range?• Reduces competitive pressure
Large vs. Small Supermarkets
One-stop vs. Top-up shopping
Out-of-town vs. Town centre location
Large supermarkets have a wider range both within and across(?) categories than a smaller supermarket.
Parking, other facilities
Price flexing and “PQRS”
Supermarkets vs. Specialists
Supermarkets internalise agglomeration effects of providing a range of products, a group of independent specialists does not (Hay and Smith, 2005)
Specialists offer expert advice Specialists have a wider within category range(?) Supermarkets have more categories Perceived higher quality
Methodology
Multiple-entrant qualitative response model
Cross-sectional data
Limited to pairwise comparisons– At present limited to comparisons where dependency
between firms is only in one direction• Eg small supermarkets profits depend on number of large
supermarkets but not vice-versa
Methodology
The profits of a particular type of supermarket, k where k = 1, 2 are given by:
– Where S is market size (population and nearby population), X is a vector of demand/cost variables that affect firm profitability within a market.
),,,,(
),,,(
212
11
NNsspecialistSf
NsspecialistSf
iii
iii
ki
ki
ki
X
X
Methodology
Assuming that large supermarkets constrain small supermarkets (but not vice versa). The latent profits are given by:
A firm of a particular type will enter market i when:
SiLiSiSiSSSiSiSSi
LiLiLiLLLiLiLLi
NsspecialistgXS
NsspecialistgXS
)(
)(
kikiki 0
Methodology
Latent profits are unobserved, but if ε are i.i.d. bivariate normal, the observed number of stores of each type, Nk (k=L,S) can be estimated using a simultaneous equation bivariate ordered probit model.
NL and NS are observed such that:
LiJ
Li
Li
Li
cifJ
ccif
cif
N
11
1211
11
2
1
SiSK
SSiS
SSi
Si
cifK
ccif
cif
N
1
21
1
2
1
Methodology
Therefore the probability of observing NL=j and NS=k in market i is:
Assuming ε are bivariate standard normal with correlation parameter ρ, this corresponds to a simultaneous equation bivariate probit model with one endogenous regressor.
),Pr(),Pr( ,1,,1, kSSkSjLLjLSL cccckNjN
Note:
There are several special cases nested within this model:
γ=0 (coefficient on endogenous regressor) – seemingly unrelated model
ρ=0, univariate probit model
Sample Markets Aggregate postcode districts (e.g. NR1, NR2) into postal towns (e.g. Norwich).
Vary in population from 142 (St. Martin’s, Isle of Scilly) to 1,015,043 (Birmingham).
Each London borough is classified as a separate postal town.
1239 postal town markets
Sample Markets - filtering Why filter?
– Isolated markets– Partial and sub-markets– Categorisation of number of firms
Cleeren et al (2007) filtering rule:– Exclude markets where pop<3,000 or pop>25,000
Bresnahan and Reiss (1991) filtering rule:– Exclude markets within 20miles of a town with pop>1,000 – Exclude markets within 100miles of a town/city with pop>100,000
My filtering rule:– Exclude markets within 20miles of a town/city with pop>150,000– Exclude markets where pop<3,000 or pop>75,000
Sample Markets - filtering
Variables All markets Sample markets
Mean S.D. Min Max Mean S.D. Min Max
POP (1000) 42.149 72.089 0.142 1,015.043 14.890 13.305 3.014 73.776
%EMPLOY 0.442 0.053 0 0.739 0.426 0.043 0.267 0.594
OVER64 0.176 0.050 0 0.404 0.208 0.049 0.045 0.384
BUTCHERS 4.385 0.224 0 138 2.746 2.438 0 14
BAKERS 4.422 8.247 0 90 2.498 2.643 0 18
GREEN 1.390 2.801 0 37 0.878 1.112 0 6
FISH 0.641 1.294 0 13 0.722 1.308 0 9
OFFLICENCE 4.431 399 0 133 0.946 4.706 0 26
LARGE - - - - 1.448 1.541 0 6
SMALL 10.213 20.802 0 338 3.373 2.965 0 13
Large supermarkets
Defined as the following fascias of the “big four”– Tesco (Supermarket and Extra formats)– ASDA (All stores)– Sainsbury (Supermarket format)– Morrisons (All stores)
Counted in a postal town if it is within 15mins drive of the centre.
Small supermarkets
Defined as:– All non-”big four” supermarkets
• I.e. Somerfield/Co-op, M&S, Waitrose, Budgens
– All non-”big four” convenience stores
– Tesco (Express and Metro formats), Sainsbury (Local and Central formats)
Counted in a postal town if in one of the postcode districts that comprise the postal town.
Specialists
Specialist stores included:– Butchers (5449)– Bakers (5505)– Offlicences (5539)– Greengrocers (1733)– Fishmongers (790)
Counted in a postal town if in one of the postcode districts that comprise the postal town.
Other Data
Market size given by population
Control variables – percentage employed, age
Instruments– For “large” supermarkets – percentage of car owners
Final specification
Profits of a large supermarket are given by:
Profits of a small supermarket are given by:
LLDLLLLLLi NSPECAGEEMPLOYPOP 321)ln(
SLSSDSSSSSS NNSPECAGEEMPLOYPOP 321)ln(
LARGE SMALL
Variable Coefficient Variable Coefficient
lnPOP 0.739 lnPOP 0.557
Employ 2.582 Employ -2.743
Over64 -0.498 Over64 -2.505
Spec1 -0.197 Spec1 -0.340
Spec2 -0.669 Spec2 -0.532
Spec3 -0.705 Spec3 -0.814
Spec4 -0.575 Spec4 -0.444
pcarowner 3.339 large -0.585
Cut1 -1.138 Cut1 -2.212
Cut2 -0.223 Cut2 -1.730
Cut3 0.456 Cut3 -1.487
Cut4 0.892 Cut4 -1.265
Cut5 1.254 Cut5 -1.132
Cut6 1.837 Cut6 -0.995
Cut7 -0.846
Cut8 -0.733
Cut9 -0.628
Cut10 -0.516
Wald test of independent equations (rho=0)
2.73
Wald test 139.73
Red, orange and yellow indicate significance at the 1%,5% and 10% level respectively.
Red indicates significant difference to the preceding cut-point at 1% level.
Conclusions
Evidence that large and small supermarket’s entry decisions are strategically linked and jointly determined.
Some weak evidence that, for large supermarket, competitive pressure increases with additional entry.
Some weak evidence that an agglomeration of three different types of specialists has an impact on the profitability of small supermarkets but no effect on large supermarkets.
Extensions
Jointly determine whether small and large supermarkets are strategically dependent (two endogenous regressors)
Endogenise specialist entry
Analyse affect of barriers to entry affecting large supermarkets
Model product range as two-dimensional product differentiation. Change order of game.
Extensions
Model product range as two-dimensional product differentiation.
Small supermarkets
Specialists
Large supermarkets
Range w
ithin categories
Range across categories