Booking open Available Pune Call Girls Shivane 6297143586 Call Hot Indian Gi...
Session 4 a chen et al discussion
1. Are Intangibles More Productive in ICT-Intensive
Industries?
Evidence from EU Countries
by Wen Chen, Thomas Niebel, Marianne Saam
Discussed by
Cecilia Jona-Lasinio
(ISTAT and LUISS Lab)
IARIW 33rd General Conference
Rotterdam, the Netherlands, August 24-30, 2014
C. Jona-Lasinio
2. Outline
• Motivation and goal of the paper
• Empirical findings
• Major remarks
• Theoretical model
• Econometric strategy
• Suggestions
C. Jona-Lasinio
3. Motivation I
• ICT is a fundamental driver of economic growth
(Jorgenson and Stiroh, (1995) and (1999); Oliner, Sichel
and Stiroh (2007)), but ICT alone does not explain the
large country-industry productivity gap between Europe
and the U.S.
• Intangible capital has emerged as a new relevant source of
productivity growth in the U.S. (Corrado, Hulten and
Sichel, 2005, 2009) and in the European economies
(Corrado, Haskel, Jona-Lasinio, Iommi, 2012 and 2014a).
• Microeconomic evidence emphasyses the relevance of
co-investments in training and organisational change to
generate ICT productivity benefits (e.g., Bresnahan,
Brynjolfsson, and Hitt, 2002; Brynjolfsson, Hitt, and Yang,
2002).
C. Jona-Lasinio
4. Motivation II
• Macroeconomic productivity studies suggest that the
returns to ICT and productivity growth are higher once
proxies for intangibles are explicitly accounted for, e.g.
Basu, Fernald, Oulton, and Srinivasan (2004).
• Based on an econometric analysis of a 10 country, 10 year
KLEMS sample of productivity growth in the EU, Corrado,
Haskel and Jona-Lasinio (2014a)1
found that productivity
in ICT-intensive industries is stronger in countries with
relatively fast-growing intangible capital, suggesting
complementarity between ICT and intangible capital.
1Revised, working paper version of paper presented at 2nd World KLEMS
conference, August, 2012, and 3rd SEEK Conference (ZEW, Manheim, April 2013)
C. Jona-Lasinio
5. Aim and main contribution of the paper
• The authors explore to what extent the contribution of
intangible capital to industry productivity growth is related
to the degree of ICT intensity at the industry level.
• Main contribution of the paper:
• Extend the analysis developed in Corrado, Haskel,
Jona-Lasino (2014) including industry measures of
intangible capital.
• Test different measures of ICT intensity
C. Jona-Lasinio
6. Empirical strategy
• They adopt a difference-in-difference approach a la Rajan
and Zingales (1998) to test the productivity impact of
intangible assets in ICT intensive industries (Corrado,
Haskel, Jona-Lasinio(2014a)).
• Data on intangibles are obtained breaking down the
INTAN-Invest Harmonized business investment by industry
(Niebel, O’Mahony and Saam, 2013).
• Sample: 10 European countries, 11 NACE Rev1.1 industries
(A to O) over the period 1995-2007.
C. Jona-Lasinio
8. Empirical model
∆(v − l)c,i,t = γ1∆(kNICT
− l)c,i,t ∗ DICT
c,i
+ γ2∆(kICT
− l)c,i,t ∗ DICT
c,i
+ γ3∆(kINT
− l)c,i,t ∗ DICT
c,i
+ βX + ωc,i + τt + c,i,t
The main assumption is that γ3 > 0 while γ1 and γ2 < 0
C. Jona-Lasinio
9. ICT intensity measures
observe that transport (I), financial intermediation (J), and business services (K) are ICT-intensive
industries according to all four measures; while agriculture (AtB), manufacturing (D), and
construction (F) are always ICT non-intensive. Mining and quarrying (C) remain below the median
for three measures.8
Figure 4.1: FOUR MEASURES OF E.U. INDUSTRY ICT INTENSITY
Since the ICT intensity might be endogenous, we follow Michaels, Natraj and Van Reenen (2014)
.02 .027 .042
.1
.13 .14
.25 .25
.29 .31
.47
0
.1.2.3.4.5.6.7.8.9
1
AtB F H D O G K I E C J
Ratio of ICT capital to labour services
.0057 .011 .014 .019 .028 .035 .038 .038 .061
.085
.11
0
.1.2.3.4.5.6.7.8.9
1
AtB F H C D G E O K I J
ICT capital share of value added
.043
.073 .085 .11 .12
.16
.21 .21
.26
.32
.37
0
.1.2.3.4.5.6.7.8.9
1
AtB C E F D H G O I K J
ICT capital share of total capital services
.025
.051 .055 .059 .073 .087
.11 .13
.2 .21
.24
0
.1.2.3.4.5.6.7.8.9
1
AtB C E F D H G O K I J
ICT capital share of total capital compensation
C. Jona-Lasinio
10. Empirical results
Table 5.1: COBB-DOUGLAS PRODUCT FUNCTION ESTIMATION
DV: ∆𝐥𝐧 (𝑽/𝑳) 𝒄,𝒊,𝒕
(1)
Two-capital inputs
(2)
INT augmented
(3)
Full Model
(4)
Full Model
OLS OLS OLS IV
NICT
(β1) 0.372***
(0.050)
0.313***
(0.048)
0.312***
(0.049)
0.312***
(0.056)
ICT
(β2) 0.087***
(0.026)
0.080***
(0.025)
0.066***
(0.024)
0.034
(0.024)
INT
(β3) 0.130***
(0.031)
0.161***
(0.034)
0.126***
(0.029)
𝑁𝐼𝐶𝑇× 𝐷 , (γ1) -0.060
(0.284)
-0.251
(0.271)
𝐼𝐶𝑇× 𝐷 , (γ2) -0.207**
(0.088)
0.086
(0.354)
𝐼𝑁𝑇× 𝐷 , (γ3) 0.340*
(0.193)
0.752***
(0.208)
Year dummies Yes Yes Yes Yes
N 1320 1320 1320 1320
Adjusted R2 0.187 0.209 0.214
Note: The output V (i.e. value-added) in Column (1) is not adjusted for the inclusion of intangible capital; whereas for column (2)-
(4), intangibles are added both as an input and an output. Hence, output V is adjusted for intangibles in these columns. All
specifications in column (1)-(4) include the country-industry-specific fixed effects. Standard errors shown in parentheses are
heteroscedastic-robust to country-industry clustering. Column (3) and (4) calculate the country-industry ICT intensity as the ratio
C. Jona-Lasinio
11. Empirical results
Figure 5.1: MARGINAL EFFECT OF INTANGIBLE CAPITAL
-.2-.1
0
.1.2.3.4.5.6.7
-20 -10 0 10 20 30 40
Demeaned_ICT_Intensity*100
Marginal Effect 95% lower band C.I.
95% upper band C.I.
C. Jona-Lasinio
12. Main contribution of the paper
• The empirical findings support the assumption that ICT
and intangible capital are complementary in the
production process (Corrado et al 2014)
• Results are robust across different measures of ICT
intensity and to alternative industry grouping criteria
C. Jona-Lasinio
13. Major remarks: research question
The research question should be better focused
• The interaction model, as it is specified, is based on the
assumption that the productivity impact of all capital
inputs depends on the degree of ICT intensity.
• But what is the theoretical motivation supporting this
assumption?
• What are the omitted effects that should be captured by
the interaction between ICT intensity and ICT capital
accumulation? How this is related to the complementary
relations between ICT and intangible capital?
• ... and what about the interaction with NON-ICT capital
(including Residential buildings)?
C. Jona-Lasinio
14. Major remarks: econometric approach
The econometric model is not robustly specified
• Interaction models should include all constitutive terms
otherwise they produce biased and inconsistent estimates
(Brambor, Clark and Golder (2005)).
∆(v − l)c,i,t = α1DICT
c,i + γ1∆(kNICT
− l)c,i,t ∗ DICT
c,i
+ γ2∆(kICT
− l)c,i,t ∗ DICT
c,i + γ3∆(kINT
− l)c,i,t ∗ DICT
c,i
+ βX + ωc,i + τt + c,i,t
• Omitting α1DICT
c,i it is as assuming that α1 is zero. That is
based on the expectation that DICT
c,i has no effect on
∆(v − l)c,i,t when ∆(kj
− l)c,i,t is zero, where
(j= ICT, NON-ICT, INT).
C. Jona-Lasinio
15. Minor remarks
• The econometrics has to be improved
• The output elasticities of ICT and intangible in cols 1 and 2 (Table
5.1) might be severely biased because of the well known endogeneity
of capital inputs. Instrumental variable estimation (including
statistical tests) has to be reported.
• The labour quality control might have a relevant effect on the size of
the interaction coefficients and has to be included in all the
specifications.
• The background literature has to be extended. Recently,
other studies have developed estimates of intangible
investment at the industry level: (Chun, Fukao, Hisa, Miyagawa
(2012), Miyagawa and Hisa (2013) and Dal Borgo, Goodridge, Haskel,
and Pesole (2013), Corrado, Haskel, Jona-Lasinio, Iommi(2014b)).
C. Jona-Lasinio
16. Suggestions
• Better focus the research question
• Check how/if the inclusion of all the interacted terms
affect the empirical findings
• Improve the econometrics
• Provide information about industry estimates of intangible
investment
C. Jona-Lasinio