Retail Market Structure and Dynamics:
A Three Country Comparison of Japan,
the UK and the US
Prepared for:
NBER/CRIW
July 16-17, 2007
Jonathan Haskel
Queen Mary, University of London, AIM, CeRiBA, CEPR and IZA
Ron S. Jarmin
Center for Economic Studies, U.S. Census Bureau
Kazuyuki Motohashi
Department of Technology Management for Innovation, School of Engineering, University of
Tokyo and Research Institute of Economy, Trade and Industry
Raffaella Sadun
Centre for Economic Performance, London School of Economics, AIM and CeRiBA
Disclaimers
US Disclaimer: The views expressed here are those of the authors and do
not necessarily reflect those of the US Census Bureau. All econometric
results have been screened to ensure the confidentiality of individual
respondents is protected.
UK notes and disclaimer: financial support for this research comes from
the UK ESRC/EPSRC Advanced Institute of Management Research,
grant number RES-331-25-0030. UK work was done at CeRiBA at the
Business Data Linking Branch at the ONS; we are grateful to all
institutions concerned for their support. This work contains statistical data
from ONS which is Crown copyright and reproduced with the permission
of the controller of HMSO and Queen's Printer for Scotland. The use of
the ONS statistical data in this work does not imply the endorsement of
the ONS in relation to the interpretation or analysis of the statistical data.
This work uses research datasets which may not exactly reproduce
National Statistics aggregates.
Motivation
Productivity growth is a function of not only
technology and other shocks, but how firms
and markets respond to these shocks.
Recent literature stresses the role of firm and
establishment turnover in reallocating
resources from less to more efficient
producers – Foster et. al. (2006), Haskel and
Sadun (2005) and Matsuura and
Motohashi(2005)
3
Motivation (cont.)
Thus, the ability of markets to reallocate
resources is crucially important to the
innovative process, productivity growth
and improved living standards
4
Why Retail?
• Growing share of the economy
• Strong recent productivity growth
• Undergoing significant structural
changes
5
Background: Labour productivity in
retail trade (GVA per person engaged)
US
UK
JA
Levels
1980
1995
2002
100
100
100
63
61
56
62
70
46
Growth rates
1980-95
1995-2002
2.2
5.4
2.0
4.3
3.0
-0.7
Source: Timmer, M and Ypma, G., (2006), “Productivity Levels in
Distributive Trades: A New ICOP Dataset for OECD Countries”, Working
paper GD83, www.ggdc.net/workpap.html, Table 12
6
20
20
19
19
19
19
19
19
19
19
19
19
03
00
97
94
91
88
85
82
79
76
73
70
Gross Valued Added per Person
Engaged (1997=100)
Labor Productivity Growth
(Source: EUKLEMS)
180
160
140
120
100
80
60
40
20
0
Japan
UK
US
Year
7
Productivity drivers in the Retail
Sector
• Information Technology
– Industry level (Jorgensen and others)
– Firm level (Doms, Jarmin and Klimek,
2004)
• Large chain retailers much more IT intensive
and have much more productivity and
employment growth
8
Productivity Drivers (cont.)
• Firm and Establishment Dynamics
– Foster, Haltiwanger and Krizan(2006)
• Contribution of net entry more important than
productivity growth at continuing establishments
• Large national retail chains play an especially important
role by opening more productive establishments
– Jarmin, Klimek and Miranda (2005)
• show that local retail markets are increasingly served by
fewer, larger firms often associated with large national
chains
9
Modest Growth of Establishments
2,000
1,500
1,000
500
0
19
61
19
65
19
69
19
73
19
77
19
81
19
85
19
89
19
93
19
97
Establishments in
thousands
Figure 2.2 Number of Retail Establishments
1958-2000
Year
Sources: Statistical Abstract of the U.S. and ow n calculations
from the LBD
Total Establishments
Single Location Establishments
Chain Establishments
10
Growing Dominance of Chains
Employment in thousands
Figure 2.3 Retail Employment at Single Location
and Chain Stores
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
1976
1979
1982
1985
1988
Year
1991
1994
1997
2000
Source: Own calculations from the LBD
Single Location
Chain Stores
11
Long run trend in composition of retail
sales
Share of Total Retail Sales for Single Units
and Large Chains (>100 Stores)
80
60
Single Unit
40
20
Large Chains
0
19
29
19
35
19
54
19
63
19
72
19
82
19
92
Percent
100
Year
12
Note on the Role of IT in Retail
• Prior slides show long run trend
towards chain dominance pre-dates
electronic computers and the Internet
– Telephone and Telegraph helped “pre-IT”
chains (need even older data)
– IT and the Internet have led to a
reshuffling of the top chains but have not
substantially altered the long run changes
in market structure
13
What’s so special about large retail
chains?
• They operate a very large number of
stores
– Just under 14% of all retail establishments
– Just over 12% of all retail establishments
are owned by firm born before 1976.
• Large and old firms account for a very
significant amount of retail job creation.
14
Large Retail Firms Cont.
• More IT investment
• Higher return on IT investment
• IT more integrated into overall business plan
(McKinsey, 2004)
• More investment in advertising / brand name
• Increasing returns on these investments
captured by opening more stores (Starbucks)
• Other intangibles?
15
Cross-Country Differences in Retail
Market Structure and Dynamics
•
•
•
•
U.S.
Japan
U.K.
Coordinated research design using
restricted access microdata for each
country.
16
Data
Japan
Retail Census
UK
Business Register
Store (establishment)
Store (establishment)
Data source
Primary data unit
Other units
Start year
End year
Longitudinal
Linkage method
Classification
Key variables
Comments
1997
2002
Industry, Region
Sales, Employment
Firm, Enterprise
Group
1998
2002
Survey ID
Industry, Region
Employment
Bus Reg provides emp
data. Survey gives
sales and other inputs.
1997 data exists but
noisy.
US
Census of Retail Trade,
Longitudinal Business
Database
Store (establishment)
Firm
1997
2002
Longitudinal IDs, Survey
IDs, Tax IDs
Industry, Region
Sales, Employment
Employment and Sales
data are from CRT and
longtidinal linkages from
LBD
17
Structure of the Retail Sector
Basic Retail Sector Characteristics - 2002
Japan
U.K
U.S.
1,298,038
334,627
1,114,637
9.84
5.64
3.94
Number of Firms
981,356
241,634
717,553
Single Unit Establishments
933,531
228,189
685,044
Multi Unit Establishments
316,682
106,438
429,593
7,972,803
2,984,376
14,647,675
Average Establishment Employment
6.14
8.92
13.14
Average Firm Employment
7.65
12.35
20.41 18
Number of Establishments
Establishments per 1000 pop
Employment
Structure Cont.
Establishments Per 1000 Population
Year
Japan
UK
U.S.
1997
11.06
4.93
4.10
2002
10.03
4.78
3.94
19
500
500
Figure 4.1. Frequency Distributions: Firms per 1000 Residents
400
2000
0
0
100
100
200
300
# of Counties
300
200
# of Counties
400
1976
0
5
10
15
Firms per 1,000 Residents
20
0
5
10
15
Firms per 1,000 Residents
20
Source: Own Calculations from LBD
20
Establishment (Store) Size
Distribution
Employment Size Distribution, reference year = 2002
Japan
US
U.K.
10th percentile
25th percentile
SU
1
2
MU
1
3
SU
0
1
MU
3
5
SU
1
1
MU
2
4
median
75th percentile
90th percentile
2
4
7
6
11
22
3
7
14
9
18
54
2
4
7
7
14
32
21
Median Chain Store Size – All Retail
(UK and US)
1990/1
US
single
local
regional
national
1998/9
4.5
8.3
10.1
10.6
1990/1
UK
single
regional
national
2002/3
5.0
11.0
11.9
13.0
1998/9
n/a
n/a
n/a
5.0
11.8
13.0
13.6
2002/3
2.0
4.5
9.2
2.0
4.5
10.1
22
Basic results on Dynamics
DHS Establishment birth and death rates (5 year)
Death Rate
Birth Rate
Death Rate
Birth Rate
Japan
US
% of Establishments
34.89% 40.85%
17.14% 40.14%
Employment weighted
28.30% 26.19%
25.24% 27.99%
UK
37.53%
35.68%
31.57%
32.78%
23
More on basic dynamics
Average Employment Size
Japan
emp of estabs year 1
emp of estabs in both years (continuers) year1
emp of estabs in both years (continuers) year2
emp of estabs in year 1 but not year 2 (deaths)
emp of establs in year 2 but not in year1 (births)
emp of estabs in year 2
US
5.02
5.47
5.63
4.55
8.26
6.32
12.51
15.59
22.90
8.22
8.94
13.14
UK
7.97
8.74
9.42
6.71
8.19
8.92
24
More systematic approach to comparing the
dynamics of the retail sector across the
three countries
We compare the cross sectional dispersion (standard
deviation) of establishment and firm growth rates.
We use the DHS growth rate measure which permits
entry and exit:
( xit ! xit ! s )
"it =
(( xit + xit ! s ) / 2)
25
Cross Sectional Dispersion in firm and
establishment growth rates
Cell based approach
– Establishment and Firm micro data for the
three countries are confidential and can’t
leave statistical offices
– We compute comparable statistics based
on the micro data for pre-defined cells.
• Mean and standard deviation of establishment
and firm employment growth rates (γ)
26
Cross Sectional Dispersion Regression
(Dependent Variable: std dev of employment growth)
Model
All
Continuing
Continuing
Establishments All Firms Establishments
firms
1.257
1.452
0.488
0.604
Intercept
0.063
0.046
0.032
0.033
0.023
-0.179
-0.03
0.049
Multi-Unit
0.015
0.019
0.007
0.014
0.38
0.217
0.007
-0.122
avgemp<2
0.039
0.046
0.024
0.042
0.203
0.122
0.02
0.006
2<=avgemp<5
0.026
0.027
0.014
0.022
0.196
0.049
0.051
-0.025
5<=avgemp<10
0.024
0.025
0.012
0.019
0.174
0.037
0.07
-0.056
10<=avgemp<25
0.022
0.022
0.011
0.016
Size Class
25<=avgemp<50
0.133
0.039
0.067
-0.047
0.024
0.023
0.012
0.017
50<=avgemp<75
0.082
-0.0002
0.048
-0.073
0.027
0.028
0.013
0.02
75<=avgemp<100
0.08
-0.002
0.034
-0.075
0.03
0.033
0.015
0.024
100<=avgemp
Japan
UK
-0.927
0.016
0.044
0.018
NA
NA
-0.061
0.012
-0.246
0.008
-0.047
0.011
NA
NA
-0.078
0.009
US
Observations
3 Digit ISIC Controls
27
351
yes
279
yes
333
yes
260
yes
How does this reallocation affect the
composition and structure of the retail
sector across different countries?
Does churning result in changing market
shares?
Or is it limited to only the smallest units?
28
Firm Size Class Transition Matrices
% of Firms
U.K.
1998 Size
(based on
employment)
Quintile
Births
1
2
3
4
5
2002 Total
Deaths
0.00%
10.08%
6.81%
3.91%
3.40%
3.91%
28.11%
2002 Size (based on employment) Quintile
1
2
3
4
10.99%
4.50%
3.09%
2.94%
8.09%
1.31%
0.91%
0.74%
4.74%
2.20%
1.75%
1.28%
0.66%
1.09%
3.19%
2.60%
0.38%
0.63%
1.48%
3.88%
0.19%
0.21%
0.45%
1.53%
22.50%
12.48%
10.87%
12.96%
5
1997 total
2.61%
24.13%
0.30%
21.42%
0.43%
17.21%
0.73%
12.17%
1.64%
11.41%
7.37%
13.66%
13.07% 100.00%
U.S.
1997 Size
(based on
sales) Quintile
Births
1
2
3
4
5
2002 Total
Deaths
0.00%
9.45%
7.36%
6.30%
5.54%
5.01%
33.65%
2002 Size (based on sales) Quintile
1
2
3
4
8.55%
7.16%
6.34%
5.61%
2.48%
0.99%
0.34%
0.16%
3.19%
1.31%
1.27%
0.33%
0.50%
1.49%
3.67%
1.44%
0.25%
0.37%
1.42%
4.56%
0.10%
0.12%
0.24%
1.17%
13.19%
13.34%
13.27%
13.28%
5
1997 total
4.55%
32.21%
0.06%
13.49%
0.10%
13.56%
0.22%
13.62%
1.42%
13.56%
6.94%
13.57%
13.28% 100.00%
29
More firm entry in the US than in the
UK, but how do the entrants do?
To see this, we calculate the fraction all entrants ending up in the
market share quintiles in 2002
UK
US
Fractions of entrants in mshare quintile in 2002
Q1
Q2
Q3
Q4
Q5
46%
19%
13%
12%
11%
27%
22%
20%
17%
14%
100%
100%
UK entrants more likely to remain at bottom
30
Firm Size Class Transition Matrices
% of Firms
U.K.
1998 Size
(based on
employme
nt)
Quintile
Births
1
2
3
4
5
2002 Total
2002 Size (based on employment) Quintile
Deaths
1
2
3
4
0 40,924
30,496
28,513
44,228
-33,895
855
4,539
5,103
7,965
-46,503
-3,258
-134
5,725
10,686
-39,209
-3,276
-3,731
-1,658
9,374
-983
-52,095
-3,713
-4,598
-6,370
-654,502
-5,128
-5,246
-8,854
-19,688
5
1997 total
468,148
612,309
8,235
-7,198
10,166
-23,318
16,386
-22,114
22,465
-45,294
401,612
-291,806
-826,204
927,012
26,404
21,326
22,459
51,582
222,579
U.S.
1997 Size
(based on
sales)
Quintile
Births
1
2
3
4
5
2002 Total
2002 Size (based on sales) Quintile
Deaths
1
2
3
4
5
1997 total
0 99,991 183,998 263,666 407,240 1,288,299 2,243,194
347 12,221
-151,718
10,187
8,246
6,894
-113,823
-215,932 -11,144
2,430 18,802
14,793
14,301
-176,750
-295,697 -11,678 -12,921
7,777
33,658
21,823
-257,038
-466,117 -12,374 -12,254 -19,047
19,583
86,002
-404,207
-2,462,301 -14,390 -17,518 -21,803
-60,297 1,802,537
-773,772
-3,591,765
50,752 155,956 259,582 423,223 3,219,856
517,604 31
Estabs (including births): Retail, 1998
180,000
160,000
a) 1 to 4
140,000
b) 5 to 9
120,000
c) 10 to 19
100,000
d) 20 to 49
80,000
e) 50 to 99
60,000
f) 100 to 249
g) 250 to 499
40,000
h) 500 to 999
j) 2500 to 49
20,000
i) 1000 to 24
g) 250 to 499
d) 20 to 49
a) 1 to 4
26-28
16-20
6-10
4
2
0
0
Firm Size
j) 2500 to 49
k) 5000 to 99
l) 10000+
Firm Age
32
Net Job Creation (including births): Retail, 1998
200,000
150,000
a) 1 to 4
b) 5 to 9
100,000
c) 10 to 19
d) 20 to 49
50,000
e) 50 to 99
f) 100 to 249
0
g) 250 to 499
h) 500 to 999
-50,000
j) 2500 to 49
i) 1000 to 24
g) 250 to 499
-100,000
a) 1 to 4
k) 5000 to 99
l) 10000+
26-28
16-20
6-10
4
2
0
d) 20 to 49
j) 2500 to 49
Firm Size
Firm Age
33
Role of Policy
• Policies that affect business dynamics can
affect the rate of innovation and productivity
growth
– Example from Haskel and Sadun (2007)
• Land use restriction led UK retail chains to open smaller
stores near the center of cities towns – reversing trend
to larger stores outside of city centers.
• Related to a slowing of TFP growth.
• Preliminary evidence from U.S. shows trend to larger
stores for U.S. chains continuing.
34
Distribution of avg. within-chain store size for NonSpecialized Retail Stores (ISIC 521) – UK and US
(employment weighted)
US
UK
90th
M edian
10th
1998/9
233
142
79
2002/3
282
152
82
90th
M edian
10th
1998/9
343
61
22
2002/3
374
43
18
35
Change in Chain Store Size in the U.K
(Source: Haskel and Sadun)
36
Table 9 Gross Output Regressions - Chain Stores Only
Dependent Variable is Log(sales)
Japan
U.K.
U.S.-1
U.S.-2
Coefficient Std. Error Coefficient Std. Error Coefficient Std. Error Coefficient Std. Error
log(N)
0.844
0.008
0.972
0.009
0.994
0.002
0.99
0.003
log(MSS)
0.177
0.01
0.081
0.014
0.009
0.002
0.017
0.003
Chain Dummies
Yes
Yes
Yes
Yes
Year Dummies
Yes
Yes
Yes
Yes
ISIC Dummies
Yes
Yes
Yes
Yes
R-Squared
0.6
0.929
0.849
0.85
Observations
32,744
7478
366,667
115,003
Notes: U.S.-1 model estimated on all available Economic Census Observations for 1977, 1982, 1987, 1992, 1997 and
2002. U.S.-2 moded estiamted on 1997 and 2002 data only.
37
Conclusions
• Significant differences in the structure and
dynamics of the retail sectors of Japan, the
UK and the US
• More churn of small firms in the US. But US
survivors grow more rapidly
• Size of entrants similar across the three
countries. Suggesting catch up by Japan
and UK?
• Policies that affect the ability of firms in
market economies to optimally respond to
technology shocks can slow innovation and
productivity growth.
38