scale without mass measuring and explaining the increasing turbulence in the us economy erik...
TRANSCRIPT
Scale without MassMeasuring and Explaining the Increasing Turbulence in the US Economy
Erik Brynjolfsson,* MIT Center for Digital Business and MIT Sloan School
(with Andrew McAfee, Michael Sorell and Feng Zhu)
The 2008 World Congress on National Accounts and Economic Performance Measures for NationsWashington, DC
May, 2008
Copyright (c) 2008 2
The Real Quantity of Computers
0
20
40
60
80
100
120
140
160
180
200
1987 19
8919
9119
9319
9519
9719
9920
0120
0320
05
Index: Year 2000 = 100
Copyright (c) 2008 3
Labor Productivity (Output Per Hour), Nonfarm Business
4.3
Dec-72Sep-73Jun-74Mar-75Dec-75Sep-76Jun-77Mar-78Dec-78Sep-79Jun-80Mar-81Dec-81Sep-82Jun-83Mar-84Dec-84Sep-85Jun-86Mar-87Dec-87Sep-88Jun-89Mar-90Dec-90Sep-91Jun-92Mar-93Dec-93Sep-94Jun-95Mar-96Dec-96Sep-97Jun-98Mar-99Dec-99Sep-00Jun-01Mar-02Dec-02Sep-03Jun-04
75
80
90
100
110
120
135
Q1 1973- Q4 1995Trend = 1.47% per year
Q1 1996-Q3 2004Trend = 3.06% per year
Index 1992 = 100
Source: BLS and Institute for International Economics
Copyright (c) 2008 4
IT and Productivity: The Data Speak
IT Stock (relative to industry average)
Productivity(relative to industry average)
Copyright (c) 2008 5
What Does IT Do?
Copyright (c) 2008 6
What Does IT Do?
1. Replicate Bits
Copyright (c) 2008 7
Copyright (c) 2008 8
What Does IT Do?
1. Replicate Bits
2. Replicate Processes
Copyright (c) 2008 9
Case Study: CVSBasic prescription fulfillment process:
• 27% of scripts encountered a problem
• 16% of customers disappointed at pick-upQuality Assurance
INVENTORY
INVENTORY
Drop -offDataEntry
ProductionQuality
AssurancePick -up
Standardscript path
Data entry
RPh
RPh
Shelves
Dr. call or
Production
Consult
areaDrop -off
Pick
-up
Quality Assurance
INVENTORY
INVENTORY
Drop -offDataEntry
ProductionQuality
AssurancePick -up
Standardscript path
Data entry
RPh
RPh
Shelves
Dr. call or
Production
Consult
areaDrop -off
Pick
-up
1 hour before pick-up1st step = drug utilization review (DUR)
2nd step = insurance check
Copyright (c) 2008 10
CVS: Scale without Mass
• Short-term results: customer satisfaction scores– Wait time satisfaction: 76 86– Overall pharmacy satisfaction: 86 91
• New process embedded in Enterprise IT (EIT)– 100% compliance
• Rapid roll-out to over 4000 retail pharmacies
Drop -offDataEntry
Insurancecheck
Production DUR QA Pick -up
While customer is present
Drop -offDataEntry
Insurancecheck
Production DUR QA Pick -up
While customer is present
New fulfillment process:
Copyright (c) 2008 11
Case Study: Cisco Mac Wiki- Over 10,000 Macintosh users at Cisco,
but no central IS support
- A few users established a wiki, where users could post tips, tricks, files, links and other content
- Example: tip for using the Linux printers which were ubiquitous at Cisco
- Many users all over world got up to speed entirely via Mac Wiki
- Thousands of small ideas from hundreds of users
Copyright (c) 2008 12
4 Stylized Facts and 3 Hypotheses1. IT makes it easier to replicate processes (not
just bits)– Aggregates formerly diffused– Diffuses best practices– Monitoring and compliance– CVS, Merrill, Wal-mart, Intel, etc.
2. Boundary of firm remains important• Bloom and Van Reenen 2005, Brynjolfsson, Hitt and
Yang, 2004, etc• B&N vs. Amazon; Schwab vs. Merrill; K-mart vs.
Wal-mart, etc.• “Standard” SAP is small fraction of total ERP
investment
3. IT discontinuity in mid 1990s– Enterprise IT (SAP R3, ERP, etc)– Internet and Intranets
4. Business Innovation continues– Alta Vista vs. Lycos vs. Yahoo vs. Google– Apple vs Microsoft– Schwab vs. Merrill Lynch
Since the mid 1990s, high IT industries should have experienced greater Concentration growthSince the mid 1990s, high IT industries should have experienced greater TurbulenceSince the mid 1990s, high IT industries should have experienced greater Performance Spread
=> More “Schumpeterian” competition throughout economy, not just high tech industries
Copyright (c) 2008 13
Data• Industry Turbulence and Concentration (Compustat)
– Revenue (SALES) and enterprise value (EV)– Turbulence: the average rank change of all firms in that industry (Comin
and Philippon 2005) – Concentration growth rate: % change in Herfindahl index (HI)
• IT Intensity of an Industry (BEA Tangible Wealth Survey)– IT capital service flow as a share of total capital service flow– Tangible Wealth Survey (1987-2004): 63 industries– Same measures as Stiroh (2002)
• Weights (BEA NIPA)– Full-time employees (FTE) – National Income and Product Accounts
Copyright (c) 2008 14
Average sales turbulence by industry category, 1960-2004
0
2
4
6
8
10
12
14
16
18
20
1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004
Average sales turbulence
High IT-using Industries
Medium IT-using Industries
Low IT-using Industries
Sales Turbulence Increased
Copyright (c) 2008 15
Average C20 index by industry category, 1965-2004
70
75
80
85
90
95
100
1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003
Average of C20 index (sales)
High IT Industries
Low IT Industries
Medium IT Industries
Concentration Increased
Copyright (c) 2008 16
Average EBITDA margin by industry category, 1960-2004
0
10
20
30
40
50
60
70
1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004
Average EBITDA margin IQR
Performance Spread (IQR) Increased
Copyright (c) 2008 17
Model 1 2 3 4 5 6
0.56*** 0.30*** 0.93*** 0.23 0.68*** 0.36(0.17) (0.08) (0.15) (0.18) (0.25) (0.64)
0.99** 0.68*** 1.20*** 0.81*** 1.16*** -0.77*(0.42) (0.24) (0.42) (0.28) (0.40) (0.47)
0.57** 0.787** 0.77***(0.27) (0.36) (0.18)
0.041*** 0.04*** 0.033*** 0.04*** 0.032*** 0.05***(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Weights yes yes yesIndustry fixed effects yes
Drop Outliers yes yes yes yes yesDrop low-density industries yes yes yes yes yes
Observations 1096 936 936 936 936 936Number of industries 61 52 52 52 52 52
R-squared 0.74 0.77 0.91 0.77 0.90 0.94
IT-intensity
Post-1996 dummy
Post-1996 dummy * IT-intensity
# of firms
Turbulence: Sales
Copyright (c) 2008 18
Concentration Growth: SalesModel 1 2 3 4 5 6
-0.165 -0.207 -0.843 -0.576 -1.853*** -0.55(0.32) (0.31) (0.52) (0.40) (0.72) (3.54)
4.685*** 3.534*** 4.335*** 3.172*** 3.601*** -9.380***(0.81) (0.82) (0.90) (0.92) (0.92) (3.47)
0.833 2.066** 6.034***(0.57) (1.02) (1.49)
0.001 0.002* 0.058*** 0.00 0.007*** -0.008*(0.00) (0.00) (0.02) (0.00) (0.00) (0.00)
Weights yes yes yesIndustry fixed effects yes
Drop Outliers yes yes yes yes yesDrop low-density industries yes yes yes yes yes
Observations 1098 954 954 954 954 954Number of industries 61 53 53 53 53 53
R-squared 0.04 0.03 0.01 0.02 0.06 0.19
IT-intensity
Post-1996 dummy
Post-1996 dummy * IT-intensity
# of firms
Copyright (c) 2008 19
Conclusions
• The improved ability of firms to replicate business processes via IT is associated not only with higher productivity but also with more Schumpeterian competition since 1996• More turbulence• More concentration• More performance heterogeneity
For more details, see http://digital.mit.edu/erik
Copyright (c) 2008 20
Robustness Checks
• Discrete vs. continuous IT intensity metric• Other operationalizations of IT intensity (IT
per FTE, IT as share of GDP)• Other aspects of Turbulence (e.g. EBITDA,
assets) • Other measures of concentration (e.g., C4
instead of HI; 1st difference in Log HI)
Copyright (c) 2008 21
Break Year• Chow-test indicates 1996 is a break year
– This finding is consistent with our replication story– 1995 and 1997 are also break years
• Use a difference-in-difference approach
0 1 2 396 96d D IT D ITβ β β β ε= + + + ⋅ + D96 equals 1 if year > 1996 and 0 otherwise
Copyright (c) 2008 22
Average yearly changes in Sales Turbulence, 1987-1996 vs. 1997-2004
Turbulence in sales, 1987-1996 vs. 1997-2004
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30 35 40
Average rank change in sales, 1987-1996
Average rank change in sales, 1997-2004
Below Median IT Intensity
Above Median IT Intensity
Copyright (c) 2008 23
Turbulence: Enterprise ValueModel 1 2 3 4 5 6
0.76*** 0.35*** 0.903*** 0.25 0.57* 3.74***(0.22) (0.13) (0.20) (0.25) (0.29) (0.72)
0.6 0.22 0.25 0.041 0.021 -2.34***(0.42) (0.30) (0.57) (0.36) (0.55) (0.53)
0.83** 1.12** 0.25(0.39) (0.44) (0.21)
0.07*** 0.074*** 0.066*** 0.07*** 0.07*** 0.10***(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Weights yes yes yesIndustry fixed effects yes
Drop Outliers yes yes yes yes yesDrop low-density industries yes yes yes yes yes
Observations 1096 936 936 936 936 936Number of industries 61 52 52 52 52 52
R-squared 0.83 0.84 .0.93 0.77 0.93 0.97
IT-intensity
Post-1996 dummy
Post-1996 dummy * IT-intensity
# of firms
Copyright (c) 2008 24
Average yearly changes in the Herfindahl Index of sales, 1987-1996 vs. 1997-2004
Average growth rate in HI of sales, 1987-1996 vs. 1997-2004
-15
-10
-5
0
5
10
15
20
25
30
-15 -10 -5 0 5 10 15 20 25 30
Average of yearly HI sales growth rate (%) , 1987-1996
Average of yearly HI sales growth rate (%) , 1997-2004
Below Median IT Intensity
Above Meidan IT Intensity
Copyright (c) 2008 25
Concentration Growth: EVModel 1 2 3 4 5 6
-0.214 -0.238 -1.272 -0.422 -2.499** 8.101(0.33) (0.30) (0.98) (0.41) (1.08) (13.87)
5.567*** 4.683** 6.210* 4.710** 5.772* -10.223(1.86) (1.90) (3.33) (1.92) (3.29) (6.37)
0.452 2.743** 5.776***(0.68) (1.28) (2.57)
0.004 0.005** 0.091 0.005** 0.01 0.01(0.00) (0.00) (0.10) (0.00) (0.01) (0.02)
Weights yes yes yesIndustry fixed effects yes
Drop Outliers yes yes yes yes yesDrop low-density industries yes yes yes yes yes
Observations 1098 972 972 972 972 972Number of industries 61 54 54 54 54 54
R-squared 0.04 0.04 0.01 0.04 0.06 0.10
IT-intensity
Post-1996 dummy
Post-1996 dummy * IT-intensity
# of firms
Copyright (c) 2008 26
Implications for Managers1. Heightened value of innovation
• Adjust recruiting, retention and incentives systems• Innovations can be big or small
Invest in technologies that encourage, aggregate, codify and/or propagate innovations
- ERP/SCM/CRM etc.- Enterprise-strength Web 2.0 and Social Networking
Design for agilityOthers?
Copyright (c) 2008 27
A Stylized Model
• Two firms, A and B, compete in n locations. Each firm has one branch in one location.
• Costs in each location are independent random variables, and are re-drawn each period.
• Assume the branch with a lower cost wins 100% market share in that location.
• Total market share of firm A is ratio of “wins” to total locations
Copyright (c) 2008 28
Random Cost Draws, Without Replication
• Market share of each firm is not far from 50% (or whatever the assumed distribution of cost differences is)
– Overall Concentration and Sales Rank does not change much from period to period
Without Replication
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10
Location
Cost
Cost of Firm A Cost of Firm B
Copyright (c) 2008 29
What if Low Cost Processes Can Be Replicated within Each Firm?
• With replication, e.g. via IT– All locations for each firm now replicate the cost structure of the
location with best cost structure within their own firm• As a result
– Winner-take-all => The industry becomes much more concentrated– It is easier to take over the leadership in the next period
• (i.e., the industry is more turbulent)
With Perfect Replication
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10
Location
Cost
Cost of Firm A Cost of Firm B