The productivity impact of ecommerce in

1

The productivity impact of e-commerce in the UK,
2001 and 2002 Evidence from Microdata
Ana Rincon, Catherine Robinson and Michela
Vecchi.

2
Introduction
ICT capital
Network
Productivity
3
Aims of the paper

• How does the Internet E-commerce affect company
  performance?
• Are the production and service sectors using
  e-commerce in the same way?
• Does the impact of e-commerce on productivity
  differ from 2001 to 2002?
• EDI versus Internet. Do we obtain consistent
  results?

4
Electronic networks listed in the E-commerce
survey

• External email
• Intranet
• Extranet
• EDI
• Interactive telephony
• Mobile technology
• Digital TV
• Internet

Other electronic networks
5
Computer networks and enterprise performance

• Existing evidence
• Rowlatt (2001)
• The Economist survey
• Clayton Criscuolo (2002)
• Improve the effectiveness of RD
• Facilitates access to wider markets
• Improves business processes
• Atrostic and Nguyen (2004)
• Criscuolo Waldron (2003)

6
E-commerce survey 2001 and 2002

• E-commerce 2001 and 2002 are very comparable. The
  only differences are in the order of the
  questions and survey design.
• Sample size increased from 9,000 businesses in
  2000 to 12,000 in 2001 and 2002.
• In 2000 business of lt10 employees were excluded,
  In 2001/2002 all businesses included.
• E-commerce 2001 and 2002 main questions

7
Connectivity measures

8
Use of Internet for placing/receiving orders
2001 and 2002
9
Network of systems 2001 and 2002
Integration indicator
10
ABI and E-Commerce

• The data used here
• The Ecommerce survey is a relatively small sample
  (only 12,000 enterprises included).
• Williams (2001) Clayton and Criscuolo (2002)
  description of the data .
• Number of enterprises

11
Modelling productivity - 1

• Gross output specification including intermediate
  materials (Baily 86, Basu and Fernald 97,
  Atrostic and Nguyen 04)
• We focus on e-buy and e-sell plus a combined
  variable, e-trade.
• We include a set of control variables
  Multiplant, Foreign ownership, Age of the
  reporting unit, Industry and Region.

12
Modelling productivity - 2
1)

a

a

a

a

a

Z
)
M
(
Ln
)
K
(
Ln
)
L
(
Ln
)
Q
(
Ln
ij
i
i
i
i
4
3
2
1
0
2)
å
å
e

g

g

a

a

a

reg
Ind
age
FO
Multi
i
r
i
7
6
5
13
Table 4 cross section results 2001, OLS
14
Table 4 cross section results 2002 , OLS
15
Table 4 cont. Impact of trading on the
Internet 2001 and 2002 - OLS
16
Selection bias and treatment

• OLS results are likely to be inconsistent because
  of the correlation between explanatory variables
  and residuals, which capture effects of all
  omitted and imperfectly measured variables.
• e.g. Correlation between the decision of trading
  on the web and able management, IT skills etc.
• Selectivity bias

We instrument the endogenous binary
variable (treatment effect estimator)
17
Treatment effect estimator - 1

• Considers effect of endogenously chosen binary
  variable on another endogenous continuous
  variable, conditional on two sets of independent
  variables.
• Two alternative estimation techniques
• 2sls (probit in the first stage, OLS in the
  second).
• Maximum likelihood estimator - just one step
  therefore more efficient.
• We estimate two regressions simultaneously

18
Treatment effect estimator - 2
1)
2)
19
Treatment effect estimator - 3The instrument set
(Z)

• Instrument relevance Need that endogenous
  variable X and instruments Z are correlated
• - Stock and Yogo (2004) test for weak
  instruments, based on the regression of
  endogenous X on all Z. It is based on comparing
  the F statistics with the critical values
  supplied in their paper.

• Instrument exogeneity Need Z to be uncorrelated
• with ei, , the error term in the equation of
  interest
• - To test this
• Calculate residual from each OLS regression
• Regress residual on all instruments controls
  and discard if coefficients are significantly
  different from zero.

20
Instruments for 2001 correlation with the
residuals

• Results of regressing the residual on the
  instruments - e-buy

• Results of regressing the residual on the
  instruments - e-sell

21
Instruments for 2002 correlation with the
residuals

• Results of regressing the residual on the
  instruments - e-buy

• Results of the residual on the instruments -
  e-sell

22
Table 5 The impact of buying on the Internet
on productivity 2001
All Sectors
Production
Services
Constant
1.596
1.497
1.783
(0.085)
(.121)
(.192)
Emp
0.271
0.225
0.286
(.015)
(.024)
(.018)
K (2001)
0.108
0.090
0.125
(.020)
(.029)
(.027)
Inter
0.623
0.675
0.594
(.021)
(.036)
(.027)
e-buy
0.084
0.125
0.074
(.044)
(.164)
(.064)
L. ratio test
2.96
0.44
1.09
(.085)
(.508)
(.029)
23
Table 5 The impact of buying on the Internet
on productivity 2002
All Sectors
Production
Services
Constant
1.753
1.265
2.133
(.094)
(.091)
(.182)
Emp
0.247
0.232
0.250
(.015)
(.020)
(.016)
K (2001)
0.100
0.073
0.104
(.0178)
(.021)
(.022)
Inter
0.644
0.703
0.630
(.018)
(.032)
(.021)
e-buy
0.090
0.406
0.122
(.046)
(.087)
(.054)
L. ratio test
2.40
13.63
4.04
(.121)
(.000)
(.044)
24
Table 5 cont The impact of trading on the
Internet on productivity 2001 and 2002-Treatment
25
Conclusions

• OLS
• In 2001 We do not find any significant impact of
  E-commerce on productivity.
• In 2002 We only find a positive impact of buying
  in the production sector (Criscuolo and Waldron,
  2003) and selling in the service sector.
• Correcting for selectivity bias
• In 2001 The effect of buying is positive and
  significant for the total sample, while e-sell is
  significant in production.
• In 2002 The coefficient for both buying and
  selling on the Internet are significantly
  positive in both production and services, with
  always a higher impact in the production sector.

26
Further work

• Look at the relationship between EDI and
  Internet.
• Extend the analysis using E-commerce for 2003, to
  provide better evidence of how the impact of
  e-commerce on productivity has changed over time.
  
• Look at the panel dimension, even though only
  large firms are represented
• Compare the UK experience with other countries
  (e.g. Germany)