Tieto- ja viestintäteknologian tuottavuusvaikutukset Suomessa

Linking micro data for the
analysis of ICT effects
Mika Maliranta, ETLA
Istat – Stat Fin Workshop, June 26th and
27th, Rome
Structure of presentation
• The importance of ICT analysis
• Methodological approaches
• Some findings about diffusion and
productivity effects of ICT (computers,
Internet and LAN) in Finnish business
• Some consideration of the data needs
The economic effects of ICT research
project
•
Initiated by The Ministry of Trade and Industry
to promote micro-level ICT research
• Conducted by Maliranta, Mika (Statistics
Finland/ETLA) & Rouvinen, Petri (Etlatieto Oy.)
• The purpose of the project:
1. Building the ICT micro panel data for the
Research Lab of Statistics Finland
2. Establishing research links
3. First-round analysis of the productivity effects
of ICT:
Maliranta, Mika – Rouvinen, Petri (2003), ‘Productivity Effects of
ICT in Finnish Business’, ETLA Discussion Papers 852 (see
www.etla.fi)
Motivation of the study
• The Finnish economy (manufacturing) has benefited from the
catching up potential
– Catching up potential has run dry by the early 90s (see Graph 1)
 Finnish economy needs a new source of productivity growth
– Are the “new economy” tools (i.e. computers, networks, etc.) the
solution?
• Services account for an increasing proportion of labor, output
and ICT use
 Diffusion of ICT to services is an important element of economic growth
 Productive use of ICT in services is crucial
 Analysis should cover service sector (the problems of ‘manucentrism‘)
• The Finnish statistical system provides us with a great
opportunity for comprehensive economic analysis
– Relatively good quality data from the Finnish ICT surveys conducted for
years 1998– Linkable comprehensive registers and other survey data
 needed for analyzing productivity effects and controlling background
factors (e.g. education)
The research question
• The research question of our ultimate interest:
The effect of ICT on aggregate (labor) productivity
 Y  X  P X
GDP

Y

• Some simple algebra:



X
i
i
i
i
i
i
i
i
i
i
 Xi
Y 
P  i 
 i   wi Pi
  Xi Xi 
 i

• The effect is composed of two elements
1. How intensive is the use of ICT (ICT diffusion), e.g. what is
the proportion of workers that use ICT in their work?
2. How productively ICT is used, on average, by the workers?
Research approaches
• Macro: the use of aggregate data
– Industry and/or country data (OECD studies etc.)
– Growth accounting (strong assumptions about the
behavior of the firms)
• Micro: the use of micro data
– Firm/plant data (difficulties in getting representative,
comprehensive and reliable data)
• Micro-micro:
– case studies (difficulties in getting general conclusion)
Data compiled @ Statistics Finland
Statistics Finland's
Internet use and
e-commerce in
enterprises -surveys
are the primary ICT
data source
('98, '99, '00, '01, '02)
Industrial
Statistics
(plant)
Manufacturing &
selected services
aggregation
Samples range
from 1300 to 2700
(leaving a few
hundred obs. for
panel analysis)
analysis
Employment
Statistics
(individual)
ICT
Survey
Financial
Statements
Statistics
A 4-page questionnaire
collects a wealth
of information (but
ICT investment &
staff not covered)
The Confederation of Industrial
Employershas its own e-business &
IT investment surveys
R&D
Survey
Innovation
Survey
Diffusion of ICT use
• ICT is a recent phenomenon, some chilling
in the diffusion in the very recent years
(see Graph 2)
• The proportion of workers equipped with a
computer has increased
– 10 percentage points in manufacturing and
– 6 percentage points in services in a few year’s
time
• Internet usage has increased more rapidly
Measuring ICT’s productivity effects
• Hypothesis:
– A worker equipped with ICT (computer, internet or LAN)
is more productive, on average, than a worker without
ICT, measured by 
• Other firm and worker characteristics need to be
controlled carefully!
• Measurement
– Labour:
– In ‘efficiency’ units:
– Production function:
Productivity effects of computers
Table. Productivity effects of ICT in Finnish businesses
 ,
 


   1
Model
ICT-variable
sample
1
COMP UTER
all
0,095
0,126
0,856
0,018
2
COMP UTER
all
0,099
0,129
0,871
0,000
3
LAN
all
0,148
0,122
0,870
0,008
4
LAN
all
0,153
0,123
0,877
0,000
5a
COMP UTER
young
0,277
0,122
0,858
0,020
5b
COMP UTER
middle
0,096
0,125
0,856
0,020
5c
COMP UTER
old
0,042
0,133
0,848
0,020
6a
LAN
young
0,234
0,084
0,908
0,008
6b
LAN
middle
0,148
0,121
0,871
0,008
6c
LAN
old
0,117
0,156
0,836
0,008

output
elasticity
of ICT
0,111
0,114
0,053
0,170
0,175
0,081
0,323
0,113
0,049
0,139
0,258
0,169
0,140
0,122
0,056
0,083
0,054
0,024
0,080
0,065
Findings
• Computers improve a worker’s productivity by 10-20 %
 consistent with the economic theory and earlier estimates
 roughly a half %-points of annual output growth can be attributed to
ICT: (10%/3 years)*15%=0.5% per year.
 Output elasticity of ICT capital is around 5 - 8 %.
• Significant differences between different ‘technologies’,
sectors and firms
– Young firms use ICT more productively than older ones
– Internet (external communication) very productive in the young service
firms and very unproductive in the old manufacturing firms
– LAN ( internal communication) quite productive in manufacturing firms
• Important to control labor characteristics and other relevant
factors;
– dropping the controls for educational levels and fields doubles the
estimates of ICT effects
Some consideration of the data needs
• Careful analysis of productivity effects of ICT calls for
good panel micro data
– Large and representative samples to obtain “degrees of
freedom” for the analysis
– Linkability with registers and other surveys
A need for co-ordination between surveys (and consideration for
respondence burden)
• Avoidance of asking the same question twice (or three times)
• A lot of various information from the same firms (in the same year)
– The needs of the panel analysis
• The same information from the same firms from the different years
• Conflict with the need sharing respondence burden through rotation
• The ‘long differences’ are more useful than the ‘short differences’
 A firm may be included in the sample, say, every second or three
years, not necessarily in the successive years
Graph 1. Catching up potential has run dry in
Finnish manufacturing, USA=100
Belgium
Canada
Finland
France
West-Germany
Japan
Netherlands
Portugal
Sweden
United Kingdom
120
100
80
60
40
20
1999
1997
1995
1993
1991
1989
1987
1985
1983
1981
1979
1977
1975
0
Source: Maliranta (1996), ICOP database, Groningen University
Back
Graph 2. Diffusion of ICT use among
the workers
The increase of the proportion of workers using ICT (computers or internet)
Services
10 %
10 %
8%
8%
percentage points
percentage points
Manufacturing
6%
4%
2%
0%
-2 %
1998-99
1999-00
2000-01
The proportion of
computer users
4,5 %
3,3 %
2,1 %
The proportion of
Internet users
6,5 %
8,8 %
4,1 %
6%
4%
2%
0%
-2 %
1998-99
1999-00
2000-01
The proportion of
computer users
2,0 %
5,8 %
-1,7 %
The proportion of
Internet users
3,6 %
6,4 %
2,0 %
Back