Product variety and technical
change
Richard Frenscha and Vitalija Gaucaite-Wittichb
Osteuropa-Institut München and Department of Economics, University
of Regenburg. Scheinerstr. 11, 81679 München, Germany. Email:
[email protected]
b Economic Commission for Europe, United Nations. Palais des Nations,
1211 Genève 10, Switzerland. Email: [email protected]
a
What is this paper about?
Measurement of technical change with product variety data
New theories of growth suggest that the variety of capital goods
(CG) used in production represents the state of technology
We embed a technology gap approach (see below) into such a
growth model, where CG variety is technology and derive a
testable hypothesis on how a variety measure should behave to
qualify as a measure of technology:
Average yearly rate of change of CG variety available for
production (i.e., technology) depends negatively on initial
variety and positively on human capital (conditional technology
convergence)
Change in variety/technology: a technology
gap approach (Nelson and Phelps, 1966)
Variety/technology frontier At expands at constant rate g
Each country’s change in variety/technology
„
„
is according to a weighted average of At and ht
is limited by human capital
h&t
= µeψu ( ht / At ) −γ
ht
Š u: time spent accumulating skills instead of working
Š ψ: returns to human capital formation
Š µ: efficiency of human capital formation, i.e. institutional
influences
„
Log-linearisation of this equation around the steady-state yields
our conditional technology convergence
Trade-based measurement of variety
Data for 46 OECD and transition countries’ trade with ROW and 55
partner countries between 1992 and 2004
Classification: SITC Rev. 3, 5-digit level, covering 3,114 items
Classification by Broad Economic Categories (BEC) allows grouping
items into three basic SNA categories: 471 capital good items, 1,899
intermediate good items and 704 consumer good items
Let product differentiation also reflect country of origin.
Then, available product variety: the number of imported items
times the number of their countries of origin plus the number of
exported items
„
Maximum count of available product variety of all items is 174,384
Variables for testing the conditional
technological convergence hypothesis
We test our hypothesis with two potential technology measures:
Š Variety of available capital goods relative to variety of available
consumer goods, i.e. CGVjt = VarCapjt/VarConjt, and
Š Variety of available intermediate goods relative to variety of available
consumer goods, i.e. IGVjt = VarIntjt/VarConjt.
„
each relative to consumer goods variety, to escape interdependence
between technology, trade, and income.
Dependent variables:
„
Average yearly rates of change of CGVjt and IGVjt
State variables:
„ Initial state: log CGVjt or log IGVjt ( - )
„
Steady state (human capital): u25j,tT: avg. years of school of
population aged 25 and over (+)
More variables for hypothesis testing
Control variables:
„
ReformMeasure_Levelj,tT × Initial state: ReformMeasure_Levelj,tT
are dummy variables indicating whether a country has within a
certain reform field made the step towards a certain level on the
EBRD scale within a given period.
„
„
We define a number of such measures to control for transition
efforts in various reform fields and interact with initial variety (-)
Also: country size effects on specialisation and specific demand
effects
Product Variety Growth Regressions
Dependent variables: Average yearly rates of change of CGVjt and IGVjt
C a p ita l
g o o ds va r iety
In t er m e d ia t e
g o o ds va r iety
E x p la n a to r y va r ia b les :
In itia l v a r iet y
lo g C G V jt
or
lo g IG V jt
– 0 .0 2 8 * * *
(– 3 .2 4 )
– 0 .0 3 7 * * *
(– 4 .0 9 )
A v er a g e y ea r s of s c h o o lin g ,
u 2 5 f j,tT
0 .0 0 1 2 * *
(2 .0 7 )
0 .0 0 0 3
(0 .4 9 )
M ic r os ta te,
M ic r o 1 × lo g C G V jt
or
M ic r o 1 × lo g IG V jt
0 .0 2 0 * * *
(3 .1 5 )
– 0 .0 5 4 * * *
(– 5 .1 9 )
In v es t m e n t-c o ns u m p t io n r a tio c ha n g e,
in v _ c o n j,t+ T – in v _ c o n jt
0 .0 7 1 * * *
(4 .5 7 )
0 .0 3 4 *
(1 .9 2 )
B a n kin g r e f o r m ,
B a n k _ 2 j,t T × lo g C G V jt
or
B a n k_ 2 j,t T × lo g IG V jt
O b s er v a tio n s (1 9 9 3 – 9 8 , 1 9 9 9 – 2 0 0 4 )
– 0 .0 6 5 * * *
(– 5 .5 9 )
6 4 (2 5 , 3 9 )
0 .0 3 9 * * *
(4 .6 1 )
6 4 (2 5 , 3 9 )
A d j. R -s qu a r ed , (1 9 9 3 – 9 8 , 1 9 9 9– 2 0 0 4 )
0 .5 9 , 0 .5 8
0 .6 1 , 0 .4 6
Notes: 3SLS estimates over two 5-year period equations using lagged initial state variables as instruments.
t-statistics in parentheses. * (**, ***): significance at 10, (5, 1) per cent. Interval dummies not reported.
Conclusions: Product Variety Growth
Capital goods variety (CGV) versus intermediate goods variety (IGV)
Trade-based count measures of available CGV behave ”as
if” representing technology; IGV measures do not
In CGV regressions:
„
„
„
conditional technological convergence among OECD and transition
countries
speed of technological convergence corresponds to that of per
capita income
(Only) Banking reforms exert a significant – and positive – effect
on speed of technological convergence
Additional slides
Available product variety by Broad Economic Categories, 2001
C on sum e r g oo ds
C a p it a l &
I n t e r m e d ia t e g o o d s
G e rm a n y
It a l y
F ra n c e
U n it e d
K in g d o m
U n it e d
S ta te s
N e t h e r la n d s
B e lg iu m
R u s s ia n
F e d e r a t io n
S p a in
A u s t r ia
C z e c h
R e p u b lic
S w it z e r la n d
S w e d e n
H u n g a ry
C a n a d a
N o rw a y
R o m a n ia
D e n m a rk
T u rke y
F in la n d
S lo v a k ia
S lo v e n ia
U k r a in e
C r o a t ia
G re e c e
B u lg a r ia
P o rtu g a l
L it h u a n ia
Ir e l a n d
E s t o n ia
S e r b ia
& M o n te n e g ro
L a t v ia
P o la n d
B e la r u s
K a z a kh s ta n
Ic e l a n d
T h e
F Y R
o f M a c e d o n ia
L u xe m b o u rg
M o ld o v a
A lb a n ia
A z e r b a ija n
G e o r g ia
A r m e n ia
T u r k m e n is t a n
K yrg yz s ta n
T a jik is t a n
0
15000
30000
P ro d u c t v a rie ty c o u n t
45000
60000
Variety, trade and technology
Interdependence between technology, trade, and income calls for
simultaneous equations approach
Short-cut: consumption goods variety is a pure trade measure, does
not influence income via a technology channel
Define two potentially technology-relevant variety measures as
„
„
the product variety of capital goods relative to the product variety of
consumer goods, i.e. CGVjt = VarCapjt/VarConjt, and
the product variety of intermediate goods relative to the product variety of
consumer goods,i.e. IGVjt = VarIntjt/VarConjt.
Both measures do not depend on trade – and thus on income – unless
there are asymmetrical effects on capital, intermediate, and
consumption variety via trade, for which we have to control.
Estimating conditional technological
convergence with panel data
„
„
After data cleaning, we have observations for up to 39 countries
between 1992 and 2004
We follow Barro and Sala-i-Martin (2004, ch. 12, per capita
income convergence testing):
Š Form two 5-year period equations and estimate as a system with
3SLS
Š Measurement bias from initial states dealt with by using lagged
initial state variables as instruments
Š Less prone to measurement bias than other panel estimators, and
appropriate when there is both heteroskedasticity and
contemporaneous correlation across periods
Sensitivity of Product Variety Growth Results (1)
Results are robust to sample composition
Results are robust to measurement of schooling
„
u25, u15f, and u15
Results are robust to the use of additional regressors
„
population density, share of urban population, mid-period logs of
the sum of residents’ and non-residents’ patent applications and
the sum of residents’ and non-residents’ patent applications per
employee, respectively
Results are robust alternative microstate definitions
„
Micro2 for a labour force of less than 1 million, and Micro3 for a
labour force of less than 2 million
Sensitivity of Product Variety Growth Results
Sample composition
With the full sample: we assume that countries are small
(taking the expansion of frontier technology as given)
“Small country” sample: excludes countries where this
assumption seems in doubt due to their prominence in
innovation activity (six G-7 countries in our sample),
thus reducing potential country heterogeneity bias
Results are robust to sample composition.
Sensitivity of Product Variety Growth Results (2)
Alternative policy reforms in transition
„
„
Only banking reform and interest rate liberalisation up to EBRD
level 2 exert a significant on speed of technological convergence
in CGV regressions.
No significant effect from large scale privatisation, first stage
reforms (i.e., liberalisation of prices and foreign trade plus small
and large scale privatisation), or governance and enterprise
restructuring.
Sensitivity of Product Variety Growth Results
Policy reform in transition
6 dummy variables from EBRD measures, ReformMeasure_leveljt:
whether a country has made the step towards some level in the EBRD
scale within given period; interact with speed of convergence
Godoy and Stiglitz (2006): no significant endogeneity problem in per
capita growth and policy reforms context. Reforms exogenous
Aghion et al. (2005) imply banking reform could have a positive
impact on technical change over and above that on trade
Results: only banking reforms exert a significant and positive effect
on speed of technological convergence in CGV regressions; slightly
increase the point estimate of estimated speed of convergence,
leaving point estimate of schooling variable unchanged (at higher
significance)
I.e., results are robust to the introduction of policy reforms in
transition with overall fit of estimations improved with banking reform