RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011 Conference - Zagreb: October 13-15, 2011...
24
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011 Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia «Learning-by-Exporting» Innovation Effects for Russian Manufacturing Firms: Evidence from Panel Data Victoria Golikova, [email protected]Ksenia Gonchar, [email protected]Boris Kuznetsov, [email protected]Institute for Industrial and Market Studies
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011 Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia « Learning-by-Exporting» Innovation
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia
Learning-by-Exporting Innovation Effects for Russian Manufacturing
Firms: Evidence from Panel Data Victoria Golikova, [email protected]
Ksenia Gonchar, [email protected] Boris Kuznetsov, [email protected]
Institute for Industrial and Market Studies
Slide 2
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Structure
of the presentation Motivation Background facts on Russias foreign
trade Research hypotheses Data description and descriptive
statistics Models and methodology Results and conclusions 2
Slide 3
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Motivation
key research questions Is there a chance for Russian manufacturing
firms to take advantage of trade liberalization and learn
globalization lessons? If yes, what would be the transmission
mechanisms? What types of firms benefit most from trade incentives?
In what aspects are learning-by-exporting effects most pronounced?
Does export destination matter? How much different are Russian
companies in their ability to learn by exporting from their
counterparts in other transition economies, who are more globally
integrated and involved?
Slide 4
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia The
underlying theoretical model is the Melitz and Bernard model for
heterogeneous firms engaged in international trade (Bernard et al,
1999 Melitz, 2003), which predicts that since more productive firms
generate higher profit gains, they are able to afford high entry
costs. This would lead to inter-firm reallocations toward more
productive firms, resulting in aggregate industry productivity
growth. Constantini and Melitz (2008)show how the market size may
affect a firms choice in favor of exports or innovations, and prove
that a firms productivity growth is endogenous, influenced by its
decision to innovate. Theoretical work has proved that the export
status and innovations are at least complementary as they provide a
potential opportunity for new knowledge (Aw et al., 2005;
Castellani and Zanfei, 2007), and also due to possible links
between product and process innovations (Damijan et al., 2008).
Empirical testing of interaction between exporting and innovations
produces mixed results (Wagner, 2007). Empirical studies utilizing
data from emerging and transition economies show that global
engagement tends to intensify innovative activities of firms
(Bustos, 2011 for Brazil-Argentina bilateral trade, Sutton, 2007
and Gorodnichenko et al, 2010, for emerging market economies). The
question of who has better chances to overcome a technology gap
firms lagging farthest behind or those closer to the leaders still
gets different answers in the literature. Economic literature on
LBE effects
Slide 5
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Some
authors believe that the bigger is the gap the better would be the
firms chances for LBE and for catching up with the leader
(Fagerberg, 1994, Julan Dua et al, 2010). Others argue that the LBE
effects are likely to be stronger for firms closer to the
technology frontier (Aghion, Bessonova, 2006). Studying LBE the
authors note varying sector-specific firm response. Julan Dua et
al, 2010, prove that exporting has virtually no effect on firm
behavior in mature low-technology sectors, while LBE effects are
more pronounced in medium- and high-technology industries.
Moreover, learning effects may not be immediately seen, coming with
a lag. Many studies find that the probability of innovative
learning-by-exporting depends on export destinations. Thus, exports
directed to high income countries require higher quality workforce
and encourage the exporter to develop business models involving
fringe distribution, transportation and publicity services
(Verhoogen, 2008, Matsuyama, 2007, Brambilla, Lederman, Porto,
2010). The Russian case of LBE effects including the impact of
export destinations (CIS and OECD), was explored by Wilhelmsson,
Kozlov (2007). They focus more on the learning outcomes, i.e.
increased productivity of exporters. The study finds that in this
sense of learning, exporting to developed countries has a more
pronounced effect for export starters. However, later on, the
differences between CIS exporters, non-exporters and OECD exporters
tends to fade out, which does not allow to make decisive
conclusions about the impact of export destination on productivity
growth.
Slide 6
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Source:
WTO. International trade statistics 2009
http://www.wto.org/english/res_e/statis_e/its2009_e/its09_trade_category_e.htm
Background: low share of manufacturing in Russian export
Slide 7
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Changes in
absolute volumes of exports in selected manufacturing industries,
in US$bn, actual prices
Slide 8
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Research
hypotheses Hypothesis 1 Exporters tend to be more innovative than
non-exporters, as they introduce new technologies and new products,
undertake/contract R&D, promote new managerial technologies and
retrain and upgrade their managerial staff. Hypothesis 2 A long
presence in export markets tends to enhance learning effects. In
other words, incumbent exporters learn quicker than export
starters. Hypothesis 3 Destination of trade (to either developed or
CIS countries) matters: exporters exclusively to CIS show weaker
learning effects than exporters to non-CIS.
Slide 9
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Data
description First round of the Survey: Conducted in Autumn 2005 for
Russian Ministry for Economic development in cooperation with the
World Bank; 1002 large and medium size firms surveyed in 8 2-digit
manufacturing sectors and in 49 regions of Russia Second round of
the Survey: Conducted in Spring 2009 for Russian Ministry for
Economic Development; 957 large and medium size firms surveyed in 8
2-digit manufacturing sectors and in 48 regions of Russia Panel -
499 firms (surveyed twice) NB: Small (less than 100 employees) and
very large companies (over 10 000 employees) were not included in
the sample
Slide 10
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia To verify
H1 and H2 we divide the sample by 4 groups of firms: Old exporters
firms which reported export both in 2005 and 2009 (NB: we presume
those forms to export continuously) Newexporters firms which
reported no export in 2005 but some in 2009 Ex-exporters firms
which reported export in 2005 but reported no export in 2009
Non-exporting firms no export reported in both rounds To verify H3
we divide the sample in three groups by destination of export in
2009: Firms with some export outside of CIS Firms exporting
exclusively to CIS Non-exporting firms Those groups are used as
dependent variables in multinominal regressions General approach
(1)
Slide 11
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia General
approach(2): determinants To estimate dependent variables, which
take discrete values of 0-1, we use standard probit regression to
estimate the dependance of an indicator in 2009 on the previous
value of the same indicator in 2005 (lagged values of dependent
variables). To avoid endogeneity issues, related to firm
size-ownership causality direction, we use lagged values of these
predictors. We use log of number of employees in 2005 to catch the
size effects We control for foreign owners and for the state as an
owner as well as to be a part of a large holding company. One
additional factor that we presume may be important is the age of a
firm by dividing them into three groups: those which existed (were
established) before 1992, created between 1992 and 1999 and the
rest of the sample. For the second model where the geographical
destination of exports is a dependent variable we use the same list
of independent variables. In both models industries are controlled
for.
Slide 12
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia
Distribution of firms by past and present export activity
Slide 13
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia 20052009
Total exporters, including: 50.757.0 -exporters to CIS countries
only 23.526.5 - exporters to the global market 27.230.5 Geography
of export flows (% of firms )
Slide 14
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Share of
organizational and managerial innovators in export status groups in
2009, %
Slide 15
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Share of
organizational and managerial innovators in groups differing by
export direction in 2009, %
Slide 16
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Dependent
variables
Slide 17
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia
Predictors
Slide 18
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia LBE
effects: Variables and Econometric approach LEf i -various mesures
describing firm activities in innovations, managerial and
organizational improvements Exp_status reflects export activity in
both rounds of the survey, Size the size of firms as measured by
the number of employees Foreign indicates a foreign shareholder
State indicates a government share in the ownership structure
Holding indicates that a firm is part of larger integrated group of
companies Age period of establishment of a firm Ind dummy variable
for 8 two-digit manufacturing industry codes T-1 indexes show the
lagged variables that we measure for the previous period of
observation. We use standard probit regression with non-exporting
firms as a reference group To estimate dependent variables, which
take discrete values of 0-1, we use standard probit regression to
estimate the dependence of an indicator in 2009 on the previous
value of the same indicator, firm export status and other firm
characteristics.
Slide 19
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Regression
results for the model estimating dependence of firm innovative
behavior on its export status LRN1 IT division LRN2 ISO certi
fication LRN3 R&D spending LRN4 managers with MBA LRN5 New
product LRN6 New technology LRN7 Domestic bench marking LRN8 Inter-
nationa bench marking LRN9 Out sourcing LRN10 Design unit LRN11
Marketing unit LRN12 After- sale service unit LRN_05_i
***1.01***0.96*0.27***0.50***0.43*0.100.18***0.30***1.21***0.88***0.84***1.11
DE_1
***0.57**0.43***0.57*0.32**0.36***0.38**0.53***1.040.33**0.340.09-0.29
DE_2 ***0.470.12*0.380.330.160.180.20***0.71**0.510.18-0.11-0.09
DE_3 -0.34-0.230.12-0.25-0.36-0.320.220.250.00*-0.56-0.18-0.30
Size05
***0.33***0.31***0.28***0.20**0.15***0.220.150.09***0.410.11**0.16***0.28
F05 -0.20-0.430.100.24-0.250.09**-0.730.03-0.18-0.110.070.22 S05
-0.06-0.03-0.06 0.14-0.230.44-0.020.250.420.100.18 Holding05
0.02-0.050.050.08-0.06 0.070.0010.23**-0.30-0.15-0.20 age1
0.310.02-0.20***-0.530.25*0.38**0.510.14*-0.40-0.11-0.050.12 age2
0.450.25dropped 0.33**0.580.430.24-0.200.15-0.110.13 age3 dropped
-0.87-0.33dropped N obs 487493456472499 R2 0.270.250.190.130.08
0.140.20.190.110.35 Note: *** - significance at 1 percent, ** - 5
percent, * - 10 percent. In groups by export status, non-exporters
(those who did not report exporting in either round of the survey)
are a reference group. LRN_05_i values of respective dependent
variables in the previous period. Industrial dummies are included
in the model but not reported in the table
Slide 20
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia To test
third hypothesis assuming LBE effects depending on CIS or non-CIS
export direction, we modify the model by replacing the export
status variables with variables indicating if the firm exports to
non-CIS, only to CIS or is not engaged in exporting at all. LBE
effects and destination of exports
Slide 21
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia LRN1 IT
division LRN2 ISO certi- fication LRN3 R&D spending LRN4
managers with MBA LRN5 New product LRN6 New technology LRN7
Domestic bench- marking LRN8 Internationa bench- marking LRN9 Out-
sourcing LRN10 Design unit LRN11 Marketing unit LRN12 After-sale
service unit
LRN_05***1.01***0.99**0.27***0.50***0.460.0980.20***0.34***1.19***0.88***0.84***1.11
Non CIS05
0.27***0.65***0.500.220.090.23**0.58***0.76-0.09**0.340.22-0.13
CIS050.21-0.020.210.03*0.260.180.28***0.500.18-0.05-0.03*-0.35
Size05***0.34***0.29***0.28***0.20**0.18***0.240.150.12***0.460.11**0.14***0.25
F05-0.13-0.480.150.27-0.160.15**-0.720.13-0.10-0.080.050.18
S05-0.06-0.02-0.05-0.090.16-0.220.45-0.010.25*0.410.090.17
Holding050.07-0.020.060.12-0.02-0.030.080.030.23*-0.26-0.13-0.20
age1-0.15-0.000.64***-0.540.240.35**0.510.09*-0.45-0.12-0.050.00
age2dropped0.200.84dropped0.31**0.550.420.22-0.190.10-0.14dropped
age3-0.42dropped -0.32dropped -0.12
ind80.12***0.88***0.74-0.210.280.20-0.050.36-0.30-0.29-0.06***1.36
_cons***-3.27***-2.97*-3.39***-1.42***-1.83***-2.48***-0.29**-1.12-***3.55-0.68***-1.48***-3.00
N obs487493456472499 R20.250.260.180.130.07
0.080.110.190.180.110.36 Impacts of export destination on
innovative behavior of firms Note: *** - significance at the 1
percent level, ** - at 5 percent, * - at 10 percent. In export
destination groups, the reference group is provided by
non-exporters, i.e. those who did not report any exporting in
either the first, or the second round of the survey. LRN_05 denotes
a lagged value of the dependent variable. Industrial dummies are
included in the model but not reported in the table
Slide 22
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia The
results obtained suggest some tentative conclusions about a
positive effect of exporting on embracement of new technologies,
primarily those in organization and management. Exporters, and,
first and foremost, long-time and continuous exporters, are more
active in monitoring their competitors, both domestically and
internationally, and more frequently engage highly qualified
managers. Exporters are more active in IT implementation. Some
evidence has been obtained in support of their higher quality
concerns, as they establish special-purpose product design units.
The most encouraging result may be seen in the evidence on
exporters higher R&D financing. Our analysis indicate that
positive changes in firm innovative behavior seem to occur
subsequently to their export entry rather than prior to it.
Moreover, this response to changes in the competitive environment
does not seem to come instantly. In other words, firms tend to
gradually learn new process and management approaches and
practices. Key conclusions
Slide 23
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia This
conclusion may be supported by the evidence that comparatively
recent export starters tend to outperform non-exporters on much
fewer parameters than the group of continuous incumbent exporters.
Moreover, learning starts from borrowing and embracement of
managerial decisions and behavior tactics of quicker returns,
including regular benchmarking, IT implementation, ISO
certification, etc. There is another conclusion that we can suggest
with some caution: non- CIS exporters are more prone to learning.
Meanwhile, firms exporting only to CIS, differ from non-exporters
mostly by their closer watching of foreign competitors. This
finding is quite consistent with other studies, specifically, with
the paper by Wilhelmsson, Kozlov (2007), which shows that
productivity gains are more likely for exporters to industrially
advanced economies. We have hardly discovered any dependence of
firm behavior on owner characteristics. This evidence is also in
line with other studies, showing that firm competitive environment
has a more significant effect on firm behavior patterns than its
ownership.
Slide 24
RESEARCH UNIVERSITY - HIGHER SCHOOL OF ECONOMICS EBES 2011
Conference - Zagreb: October 13-15, 2011 Zagreb, Croatia Thank you
for your attention