WP-2007-006
Investigating Product Cycles Using Indian Import Data
S Chandrasekhar, Abhiroop Mukhopadhyay, Rajendra R Vaidya
Indira Gandhi Institute of Development Research, Mumbai
May 2007
2
Investigating Product Cycles Using Indian Import Data
1
S Chandrasekhar
Indira Gandhi Institute of Development Research (IGIDR)
General Arun Kumar Vaidya Marg
Goregaon (E), Mumbai-400065, INDIA
Email: chandra@igidr.ac.in
Abhiroop Mukhopadhyay
Planning Unit, Indian Statistical Institute (Delhi Centre)
7 SJS Marg, New Delhi-110016, INDIA
Email: abhiroop@isid.ac.in
Rajendra R Vaidya
Indira Gandhi Institute of Development Research (IGIDR)
General Arun Kumar Vaidya Marg
Goregaon (E), Mumbai-400065, INDIA
Email: vaidya@igidr.ac.in
Abstract
We derive country ranks using disaggregated Indian import data over 1991-2005 using the
intuition that developed countries would export more advanced goods to India earlier than
other countries. We find that, consistent with theory, the degree of innovation is a significant
determinant of our ranks.
JEL code: F10
Keywords: Empirical; Data; Disaggregate; Product Cycle
1
We are grateful to Priyodorshi Banerjee and Satya P. Das for useful comments on an earlier draft.
3
Investigating Product Cycles Using Indian Import Data
S Chandrasekhar, Abhiroop Mukhopadhyay, Rajendra R Vaidya
Introduction
The product cycle theory of international trade implies an ordering of the
sophistication of goods exported by countries. Using data on exports by rest of the world to
the United States of America, for the period, 1972-94, Feenstra and Rose (2000) (F&R
henceforth) propose a methodology to rank commodities and countries.
The ranking of countries is based on the following intuition. Countries exporting more
sophisticated goods are considered more advanced. Alternatively, given two countries, the
one exporting earlier is ranked more advanced.
F&R find the country ranks consistent with theoretical predictions. Would one
generate similar rankings using data on imports by a different country given recent trade
patterns?
Apart from the fact that disaggregated (6-digit) import data (India Trades database)
are available, India presents itself as an ideal candidate for such an exercise since its import
patterns fit the model. India’s imports increased over the period 1991-2005 (Figure 1). India
imported 5248 distinct commodities
2
from 230 countries mirroring the export pattern of
countries at various stages of development. The number of commodities banned by India
have been far and few.
We find that the degree of innovation is a significant determinant of our rank
ordering. In terms of rankings, while India’s neighbours have higher than expected ranks, one
significant departure from F&R is the rise of China.
Empirical Model
Kendall and Dickinson (1990) established the procedure for ranking countries for a
balanced panel, i.e. if every country exported all commodities. However, not all goods are
exported by all countries implying that the data are censored and the panel is unbalanced. A
2
We observe 1,090,747 country commodity pairs in the data.
4
country may be too advanced to export the good during the sample period. Alternatively, it
may not be advanced enough to export a particular good during the period, but could do so in
the future.
F&R generalise the method for an unbalanced panel. Following F&R, we use the year
a country first exported the commodity to India (during our sample period) to generate two
sets of ranks: Goods Based Ranks (GBR) and Country Based Ranks (CBR).
Goods exported to India earlier are considered less advanced than goods exported
later. Countries exporting more advanced goods are ranked more advanced (GBR).
Alternatively, for each commodity, a country exporting to India earlier is deemed more
advanced (CBR). Apriori, there is no mathematical reason to expect that GBR and CBR
would be identical.
We now discuss the derivation of GBR. Let G be the set of N commodities exported
to India, G
k
the set of N
k
commodities exported by country k in the sample period and M the
set of exporting countries. Let
)(GX
i
be the true rank of goodi . For each country k, we rank
good
k
Gi ∈ by the first year that it was exported
3
. Let this rank be
ik
x . Since many factors
drive trading patterns, )(
GX
i
and
ik
x need not be identical. Let
k
N
ρ
be the number of goods
for which
iki
xGX =)( . Moreover, for countryk , we do not have rank of the goods not
exported by it. Let
(
)
min
,.....2,1
k
x
be the set of goods too primitive to be produced by country
k and (
),...1,
maxmax
Nxx
kk
+ the set of goods too sophisticated to be produced during the period.
Hence, for country
k
,
1
minmax
++=
kkk
Nxx
.
If
min
k
x were known, we could have inflated the actual rank
ik
x by
min
k
x to calculate
what would have been the ranking of goods had we observed the unsophisticated products
4
.
The crux of the empirical exercise is to estimate
min
k
x in order to calculate )(GX
i
. F&R
establish that
)(GX
i
can be derived by an iterative procedure where the initial estimate of
)(GX
i
is given by the average of
ik
x , for all
k
Gi
∈
. The parameters
ρ
and
min
k
x are estimated
from the following least square dummy variable fixed effects weighted regression
5
ikikkik
N
GXx
N
Gx
ερ
+
⎥
⎦
⎤
⎢
⎣
⎡
+
−+−=
⎥
⎦
⎤
⎢
⎣
⎡
+
−
2
)1(
)(
2
)1(
)(
min
, MkGi
k
,...,1, =
∈
3
Analogously, in case of CBR, for each good, we rank countries in the order in which they exported the good.
4
The assumption is that there are no commodities missing in the middle of the rankings.
5
The weights are given by the number of countries exporting a commodity during the sample period.
5
Next, inflate
ik
x by
min
k
x and update )(GX
i
by recalculating the average over the
updated
ik
x , for all
k
Gi ∈ . We repeat the procedure till )(GX
i
converges. Using )(GX
i
we
rank countries by the average sophistication of goods exported by them. Countries exporting
more sophisticated goods are ranked more advanced.
Country Rankings and Macroeconomic Indicators
The rankings
6
are reported in Table 1. The GBR has a correlation of 0.5 with those by
F&R. We investigate whether the country ranks, as suggested by theory, are related to
measures of innovation like ratio of expenditure on research and development (R&D) to
gross domestic product (GDP). We source data on R&D expenditure from UNDP–CDROM
(Fifteen Years of HDR 1990-2004). Using the data for the most recent year available during
the period 1990-2004, we regress the country ranks on R&D-GDP ratio and a distance
variable
7
to proxy for transport costs. We find that countries with a higher R&D-GDP ratio
are ranked as more advanced countries
8
.
We now turn to a discussion of some interesting outliers. India’s neighbouring
countries are ranked higher than expected. Their ranks are driven by two reasons: preferential
free trade agreements, and Indian firms with operations in these countries and exporting to
India. In every year, the number of goods exported to India by its neighbours is higher than
the median number of goods exported by all countries. In particular, despite not having a well
developed manufacturing sector, Nepal is ranked fourth (Mfg. GBR). This suggests inflow of
manufacturing goods from a third country through Nepal stemming from an inability to
enforce domestic content requirements.
The case of China illustrates why GBR and CBR need not be identical
9
. For most
goods China was a late entrant to Indian markets. But when China entered, it exported
sophisticated goods. In contrast, USA and other OECD countries have exported to India for a
long time, hence their high CBR.
6
After dropping countries trading infrequently, we have observations on 184 countries.
7
Source: www.cepii.fr.
8
Our results are robust to alternate specifications where instead of R&D-GDP ratio we used Hall and Jones
measure of productivity for 1988, GDP per capita for 1990 and 2004. The regression results are along expected
lines and in these specifications the distance variable is also significant.
9
The correlation between CBR and GBR, and between manufacturing GBR and CBR are 0.69 and 0.89
respectively.
