The Missing Globalization Puzzle: Another Explanation

Abstract

This study suggests another explanation of the "missing globalization puzzle" typically observed in the empirical gravity models. In contrast to the previous research that focused on aggregated trade flows, we employ the trade flows in manufacturing products broken down by 25 three-digit ISIC Rev.2 categories. We estimate the distance coefficient using the log-linear specification of the standard as well as the generalized gravity equations. Our data set comprises trade flows for 22 OECD countries that span the time period from 1970 till 2000. We observe a substantial decline in the value of the distance elasticity in most manufacturing industries.

Full text

Boriss
Siliverstovs
Dieter Schumacher
The Missing Globalization Puzzle:
Another Explanation
Discussion Papers
Berlin, October 2007

Opinions expressed in this paper are those of the author and do not necessarily reflect
views of the institute.
IMPRESSUM
© DIW Berlin, 2007
DIW Berlin
German Institute for Economic Research
Mohrenstr. 58
10117 Berlin
Tel. +49 (30) 897 89-0
Fax +49 (30) 897 89-200
http://www.diw.de
ISSN print edition 1433-0210
ISSN electronic edition 1619-4535
Available for free downloading from the DIW Berlin website.

The Missing Globalization Puzzle: Another Explanation§
Boriss Siliverstovs ∗
DIW Berlin
Dieter Schumacher∗∗
Europe University Viadrina / DIW Berlin
October 13, 2007
Abstract
This study suggests another explanation of the ”missing globalization puzzle” typically observed in
the empirical gravity models. In contrast to the previous research that focused on aggregated trade
flows, we employ the trade flows in manufacturing products broken down by 25 three-digit ISIC Rev.2
categories. We estimate the distance coefficient using the log-linear specification of the standard as well
as the generalized gravity equations. Our data set comprises trade flows for 22 OECD countries that
span the time period from 1970 till 2000. We observe a substantial decline in the value of the distance
elasticity in most manufacturing industries.
Keywords: Gravity model, missing globalization puzzle, distance coefficient
JEL code: F12.
§The paper has benefited from comments by K. A. Kholodilin and by the participants at the 9th Annual Conference of the
European Trade Study Group (ETSG), Athens, Greece.
∗DIW Berlin, Mohrenstraße 58, D–10117 Berlin, Germany, e-mail: bsiliverstovs@diw.de
∗∗DIW Berlin, Mohrenstraße 58, D–10117 Berlin, Germany, e-mail: dschumacher@diw.de

1 Introduction
The ”missing globalization puzzle” or ”distance puzzle” is well established in the traditional literature on
empirical applications of the gravity model. Coe et al. (2002, 2007) argue that the standard gravity models
that are usually estimated in the log-linear form are unable to capture the significant decline in the trade
costs brought by globalization of the world economy. In particular, they point out that even so these gravity
models based on cross-sectional regressions are able to explain the pattern of international trade relatively
well, the magnitude of the estimated distance coefficient remains stable over time. Assuming that namely
the magnitude of the distance coefficient serves as a proxy for trade-related costs, the reported stability of
the distance coefficient remains a puzzling and counterintuitive result. Coe et al. (2002, p. 3) conclude that
“globalization is everywhere but in estimated gravity models”. This fact prompts the authors to talk about
the “missing globalization puzzle”. Thus, they confirm Leamer and Levinsohn (1995) who concluded that,
contrary to popular notions of globalization, the world is not “getting smaller”.
Indeed, Coe et al. (2007) base their argument on numerous studies (Frankel, Stein, and Wei, 1997;
Eichengreen and Irwin, 1998; Helliwell, 1998; Frankel and Rose, 2000; Soloaga and Winters, 2001; Brun,
Carr`ere, Guillaumont, and de Melo, 2005, inter alia), where results of the estimated gravity models are
compared for different time periods with the typical conclusion that the distance coefficient varies in the
interval between -0.5 and -1.0 or higher and, what is more important, its value reveals no tendency to
decline over time. Hence, this conclusion on the stability of the distance coefficient seems to be robust across
different sample sizes and regression specifications.
There are several theories in the literature that explain the apparent stability of the distance coefficient
over time. (Coe et al., 2002, p. 6) mention the following four types of possible explanations of the missing
globalization puzzle: “the decline in average costs relative to marginal costs of trade over time; the increased
dispersion of economic activity; the changing composition of trade; and the importance of relative rather
than absolute costs in determining bilateral trade”. Coe et al. (2002) also provide in-depth discussion of the
proposed explanations for this puzzle. Next, (Brun et al., 2005) argue that the observed puzzle may be due
to misspecification of the transport cost function in the standard gravity models. Finally, Buch et al. (2004,
p. 297) argue that stability typically observable over time in the distance coefficient is not that surprising
because “interpretation of distance coefficients as indicators of a change in distance costs is misleading”.
Besides the theoretical considerations, there is a number of studies that argue that the problem of zero
observations that is inherent in the log-linear estimation approach of the gravity models and, especially,
various ad hoc methods used in the literature to solve this problem may have created “the missing global-
ization puzzle”. For example, Coe et al. (2007) suggest to solve the missing globalization puzzle empirically
by reconsidering the estimation method of the parameters of the gravity model. In particular, they pro-
pose to dispense with the (historically most popular) log-linear form of the gravity equation and to directly
consider the nonlinear specification of the gravity equation. Indeed, using the nonlinear specification of the
gravity model the authors show that the distance coefficient value shows trendwise decrease over time. At
the same time Coe et al. (2007, p. 36) conclude that their results “also confirm that the standard log-linear
1

specification does not yield evidence of globalization”. Similarly, Felbermayr and Kohler (2006) show that
applying a Tobit estimation of the gravity equation may resolve the distance puzzle. Dissecting trade growth
after World War II into growth of already established trade relations and establishing new trade between
countries that have not traded with each other in the past they find that distance plays an ever decreasing
role over time. This, however, has to be contrasted with the estimation of the gravity model in the log-linear
form where such decline in the value of the distance coefficient was not noticeable.
In this paper, we suggest another solution to the “missing globalization puzzle” in the gravity equation.
In this respect we would like to point out that Coe et al. (2007) and the rest of the articles cited above
estimate the gravity models using aggregated trade data. On the contrary, in our paper we employ the
trade flows at different levels of disaggregation: (i) for all products combined, (ii) for agriculture, mining
and quarrying, and manufacturing products as a whole as well as (iii) for manufacturing products broken
down by 25 three-digit ISIC Rev.2 industries. The yearly data are collected for the 22 OECD countries and
encompass the time period from 1970 till 2000.
Initially, we base our estimation results on the gravity model specified in the log-linear form in its most
basic form. Next, we estimate the generalized gravity model of Bergstrand (1989) by augmenting the basic
gravity equation by the relative factor endowment of the exporting country and the per capita income of
the importing country. The influential study of Bergstrand (1989) provides the theoretical foundation for
the gravity equation applied to disaggregated trade flows. As we apply the traditional estimation method
of the gravity equation, we unavoidably face the problem of zero observations which we solve in the natural
way by substituting them with the smallest value observed. In this way we keep all the observations in our
sample.
Our main finding is that when estimating the gravity model parameters, using trade flows broken down
by 25 three-digit ISIC Rev.2 industries, the often observed result in the models estimated using aggregated
trade flows that the distance coefficient is stable over time does not generally hold. In a large number of
manufacturing industries we find a trendwise change such that the (absolute) value of the distance elasticity is
up to 45 percent smaller in 2000 than 1970. At the same time, our estimation results obtained for the gravity
models estimated for all products combined as well as separately for agriculture and for all manufacturing
products suggest that the (absolute) magnitude of the distance coefficient remains rather stable over time.
On the contrary, we find that for mining and quarrying it substantially increases. Thus, our findings conform
with the results reported in other studies that estimate the log-linear gravity models using aggregated trade
flows.
2 Data
In the empirical analysis, for the dependent variable we employ the annual trade flows of the years from
1970 till 2000 (in US $ million) for all products combined, agriculture, mining and quarrying, manufacturing
products as a whole and broken down by 25 three-digit ISIC Rev.2 industries among 22 OECD countries.1
1Member countries in 1993, excluding Iceland and taking Belgium/Luxembourg together.
2

For this purpose the OECD foreign trade figures are appropriately re-coded from the original SITC categories.
The data on GNP (in US $ million) are taken from World Bank publications. The distance Dij (in miles)
between the countries iand jis calculated as the shortest line between their economic centres ECiand
ECjby latitudinal and longitudinal position.2The dummy variables cover: adjacency, Adjij, membership
in a preference area: European Union, EUij , European Free Trade Agreement, EF T Aij , the Free Trade
Agreement between EU and EFTA, EU −EF T Aij , the North-American Free Trade Agreement, N AF T Aij,
and Asia-Pacific Economic Co-operation, AP EC ij, in order to capture effects of regional trade liberalisation,
ties by language, Lanij, and colonial-historical ties, C olij . The value of the dummy variable is 1, if the two
countries iand jhave a common land border, belong to the respective preference zone considering the
changes over time according to membership, or have the same language or historical ties.3Otherwise the
value of the dummy variables is zero.
3 Model Specification
Our baseline specification of the gravity model in the log-linear form reads as follows
ln(Xa
ij ) = β0+β1ln(Yi) + β3ln(Yj) + β5ln Dij +γ0DU Mij +ηij ,(1)
where Xa
ij denotes the trade flows in the respective ISIC category from a country ito a country j, the variables
Yiand Yjdenote the GNP of the corresponding countries, and DU Mij = (Adjij, EUij , EF T A ij, EU −
EF T Aij , N AF T Aij, AP E Cij, Lanij , C olij)0is the vector of dummy variables as defined above in Section 2.
In sequel, we refer to the model in equation (1) as OLS2.
We check the robustness of our estimation results using the generalized gravity equation of Bergstrand
(1989) in the following form:
ln(Xa
ij ) = β0+β1ln(Yi) + β2ln µYi
Pi¶+β3ln(Yj) + β4ln µYj
Pj¶+β5ln Dij +γ0DU Mij +ηij ,(2)
where Piand Pjare the population of the exporting and importing countries, respectively. The per capita
income of country iis a proxy of the capital-labour endowment ratio of the exporting country, the per capita
income of country jrepresents the import demand conditions of the importing country. We refer to the
model in equation (2) as OLS4.
Given the available sample of yearly data that covers the period from 1970 till 2000, we estimate equations
(1) and (2) – where for simplicity the time index is omitted – for every year t= 1970,1971, ..., 2000 using the
OLS procedure. This gives us a time series of 31 cross-sectional estimates of each coefficient. However, since
our main concern is investigation of the “missing globalization puzzle”, we will focus only on the analysis of
the time pattern of the values of the estimated distance coefficient b
βt
5for t= 1970,1971, ..., 2000. The next
2The national capitals were taken as the economic centre (EC) except for Canada (Montreal), the United States (Kansas
City as a geographical compromise between the centres of the East and West Coasts), Australia (Sydney), and West Germany
(Frankfurt/Main). The formulae are: cosDij = sin ϕi∗sin ϕj+ cos ϕi∗cos ϕj∗cos(λj−λi) and Dij = arccos(cos Dij )∗3962.07
miles for ECi= (ϕi;λi) and ECj= (ϕj;λj) with ϕ= latitude, λ= longitude.
30.5 for second languages and 0.5 for historical ties until 1914.
3

section presents estimation results.
4 Results
4.1 Aggregated trade flows
In this subsection we discuss the estimation results of the distance elasticity obtained for aggregated trade
flows collected for all products combined (0) as well as for the one-digit ISIC industries such as agriculture
(1), mining and quarrying (2), and manufacturing products as a whole (3)4. Figures 1 and 3 display the
sequence of the estimated coefficients of interest using equations (1) and (2), respectively.
Observe that in both figures the estimated distance coefficient more or less fluctuates around the same
level for all products combined (0), for agriculture (1), and for manufacturing products as a whole (3) whereas
for mining and quarrying (2) it even substantially increases over time in the absolute value. Thus, our results
obtained for the aggregated trade flows further support the evidence that favors the “missing globalization
puzzle” and thus conform with the bulk of the previous literature that investigated this question.
4.2 Disaggregated trade flows in manufacturing
Next, we describe the estimation results obtained for disaggregated trade flows in manufacturing at the three-
digit ISIC level. Figures 2 and 4 display the sequence of the estimated distance elasticity using equations
(1) and (2), respectively.
First, observe that the distance elasticity estimated using either equation (1) or (2) is very similar.
Second, despite some year-to-year fluctuations it is rather safe to conclude that for most manufacturing
industries the magnitude of the distance elasticity seems to decline in the absolute value which implies that
over the observation period from 1970 till 2000 the role of distance has (substantially) decreased. This is
the main finding of our paper and, in this respect, we would like to emphasize that this result is based on
the estimation of the standard log-linear specification of the gravity model in its most basic form and it also
holds when we estimate the generalized gravity equation of Bergstrand (1989).
Our results obtained for disaggregated trade flows in manufacturing products suggest that by estimating
the gravity equations using aggregated trade flows – either for all products combined or only for manufac-
turing as a whole – one overlooks the crucial information on the time evolution of the distance elasticity
contained in the disaggregated trade flows.
Furthermore, it is of interest to quantify changes observed in the distance elasticity. In order to smooth
out year-to-year variation in the estimated distance elasticity and for the sake of robustness check we calculate
the absolute and relative change in the parameter of interest in the following two ways. First, we use the
auxiliary regressions where we regress a time series of values of the estimated distance elasticity obtained
for every ISIC category on a constant and a linear deterministic trend. The corresponding fitted values
from such auxiliary regressions are reported for each product category in the respective graphic. Then, we
4The corresponding ISIC number is given in parentheses.
4

compare the predicted values from this regression for 1970 (the initial year in our sample) and for 2000 (the
final year in our sample). Second, we compare the average value of the actual values of the distance elasticity
computed for the first three years (1970–1972) and for the last three years (1998–2000) in our sample.
Tables 1 and 2 present the results obtained by comparing the predicted values from the auxiliary regres-
sions and the averaged actual values for the first and the last three years in our sample, respectively. Each
table contains the estimated value of the distance elasticity in the beginning and in the end of the period
(columns initial and last). The columns absolute change and relative change contain the absolute and the
relative differences between the numbers that are present in the initial and last columns, respectively.
Comparison of the relative change in the distance elasticity reveals that the results are robust with
respect to the specification of the gravity model as well as to the calculation method. As seen from Tables
1 and 2, in the following industries we observe substantial decline in the value of the distance coefficient
such that its (absolute) value is up to 45% smaller in 2000 than that in 1970. These industries include food,
beverages, and tobacco (31), leather and leather products (323), footwear (324), wood and wood products
(331), furniture (332), paper and paper products (341), printing and publishing (342), industrial chemicals
(351), other chemical products (352), rubber products (355), plastic products (356), glass and glass products
(362), structural clay products (369), fabricated metal products (381), machinery (382), electrical machinery
(383), transport equipment (384), measuring, photo, and optical equipment (385), and other manufacturing
(390).
It is worthwhile mentioning that only for one industry – textiles (321) – we find that the distance elasticity
has substantially increased in the absolute value and this finding is robust regarding the estimated model
and calculation of the relative difference between the beginning and the end of the observation period. The
likely reason is the huge transfer of textile production from the OECD countries to the other – mainly
developing – countries, which are not included in our sample, that took place in the course of the period
under consideration.
For the remaining five industries which include wearing apparel (322), petroleum refineries (353-4),
pottery and china (361), iron and steel (371), and basic non-ferrous metals (372), we find that the estimated
distance elasticity exhibits neither strong nor robust evidence of change in either direction and hence we
conclude that for these industries it remains more or less the same over the observation period. Observe
that this group of industries is intensive in natural resources and at least three of them are closely related
to the mining and quarrying (2) where we find ever increasing role of distance, as discussed above.
In order to demonstrate the close correspondence between estimation results obtained for different speci-
fications of the gravity equations (OLS2 vs OLS4 ) as well as different computation method of the magnitude
of relative change (the approach based on the auxiliary regression aux vs the averaging approach ave) we
show the corresponding cross-plots of the estimated relative change observed for each industry in Figure
5. For example, the upper left graph in Figure 5 displays the relative change in the estimated distance
elasticity computed by the averaging approach for the basic gravity model (OLS2ave) plotted against that
for the generalized gravity model (OLS4ave), etc. It is worthwhile noting that the robustness of the results
is also supported by the high values of the correlation coefficient ˆρbetween the relative changes computed
5

in different ways. This correlation coefficient lies in the interval between 0.94 and 0.97.
4.3 Declining role of distance and changing trade composition
We calculated the share of each manufacturing industry in total trade of all manufacturing products observed
for each year in our sample, see Figure 6. First, we observe that there are substantial differences in the shares
of each industry in the total trade volume. For example, trade in industries such as machinery (382), electrical
machinery (383), transport equipment (384) each comprise more than 10% in total trade of manufacturing
products. On the other hand, there are industries where the corresponding trade volume is less than or
about 1% of total trade, e.g., leather and leather products (323), footwear (324), furniture (332), printing
and publishing (342), rubber products (355), plastic products (356), pottery and china (361), glass and glass
products (362), and structural clay products (369). Second, we also observe that the composition of trade
changed over time. There are industries whose share in total trade significantly increased over time, e.g.,
other chemical products (352), electrical machinery (383), transport equipment (384), measuring, photo, and
optical equipment (385), and there are industries whose share substantially decreased over time, e.g., food,
beverages, and tobacco (31), textiles (321), wearing apparel (322), iron and steel (371), and basic non-ferrous
metals (372), etc. Table 3 quantifies absolute and relative changes in the trade share of each manufacturing
industry.
In order to investigate the relationship between the declining role of distance observed in certain industries
and the changing composition in intra-OECD trade, we made cross-plots of the relative change in the distance
coefficient (calculated for the generalized gravity model(2) using the averaging approach) and the relative and
absolute changes in trade shares as reported in Table 3. Figures 7 and 8 display the respective cross-plots.
On the one hand, we find that for those five industries – wearing apparel (322), petroleum refineries (353-4),
pottery and china (361), iron and steel (371), and basic non-ferrous metals (372) – for which we observe very
little change in the distance coefficient over time as well as for the textile industry (321) we also observe a
substantial decline in their respective shares in total trade volume, measured both in relative and absolute
terms. On the other hand, those industries like other chemical products (352), electrical machinery (383),
transport equipment (384), measuring, photo, and optical equipment (385) for which we find the largest
decrease in the absolute value of distance elasticity also experienced a substantial rise in their shares in total
trade measured either in relative or absolute terms. The corresponding correlation coefficients inferred from
Figures 7 and 8 are 0.61 and 0.69, respectively.
5 Conclusions
In this study, we investigate whether the phenomenon of non-decreasing distance elasticity, labeled as the
“missing globalization puzzle” in Coe et al. (2002, 2007), that typically is found in the gravity models
estimated for aggregated trade flows, also holds for (manufacturing) trade flows disaggregated at the three-
digit ISIC level. For this purpose, we employ a data set that covers international trade flows among 22
OECD countries from 1970 till 2000. Using this data set we estimate the standard gravity model in its most
6

basic form as well as the generalized gravity model of Bergstrand (1989).
Our findings are twofold. First, when we estimate the gravity model using aggregated trade flows for
all goods combined, for agriculture, for mining and quarrying, and for manufacturing products as a whole
we find no signs that the distance elasticity declines over time in the absolute value. Moreover, for mining
and quarrying industry we find rather strong evidence that the absolute value of the distance elasticity
has increased over time. Thus, our results based on aggregated data seem to conform with the rest of the
literature on the persistence of the missing globalization puzzle. Observe that our results are based on a
more homogenous sample of the developed OECD countries and in this respect complements the rest of
the relevant studies that use more heterogenous samples of countries including both developed as well as
developing ones.
Second, when we consider manufacturing trade flows disaggregated at the three-digit ISIC level we found
that for 19 out of 25 categories the distance elasticity in 2000 has declined up to 45% compared with its
value obtained in 1970. At the same time, we find that only for one industry (textile) the distance elasticity
value has substantially increased. For the remaining five industries, we find no robust evidence that the
distance elasticity has changed over the sample period.
Our results obtained for the disaggregated trade flows in manufacturing products suggest that by esti-
mating the gravity equations using aggregated trade flows – either for all products combined or only for
manufacturing as a whole – one overlooks the crucial information on the time evolution of the distance
elasticity contained in the disaggregated trade flows. Hence, the aggregation issue seems to be relevant for
explaining the missing globalization puzzle typically observed in standard gravity models in addition to the
number of explanations that already have been put forward in the previous literature.
We also find that the manufacturing industries for which we observe the largest decline in the value of
distance elasticity are those whose share in total intra-OECD trade substantially increased over time. On
the contrary, those industries for which we find no evidence of a declining role of distance are those whose
share in total trade decreased over the three decades. Thus, our analysis suggests that transportation costs
decreased in particular for long distances in the most important and dynamic manufacturing industries.
References
Bergstrand, J. H. (1989). The generalized gravity equation, monopolistic competition, and the factor pro-
portions theory in international trade. The Review of Economics and Statistics 71 (1), 143–153.
Brun, J.-F., C. Carr`ere, P. Guillaumont, and J. de Melo (2005). Has distance died? Evidence from a panel
gravity model. World Bank Economic Review 19 (1), 99–120.
Buch, C. M., J. Kleinert, and F. Toubal (2004). The distance puzzle: On the interpretation of the distance
coefficient in gravity equations. Economics Letters 83 (3), 293–298.
Coe, D. T., A. Subramanian, and N. T. Tamirisa (2007). The missing globalization puzzle: Evidence of the
declining importance of distance. IMF Staff Papers 54 (1), 34–58.
7

Coe, D. T., A. Subramanian, N. T. Tamirisa, and R. Bhavnani (2002). The missing globalization puzzle.
IMF Working Papers 02/171, International Monetary Fund.
Eichengreen, B. and D. A. Irwin (1998). The role of history in bilateral trade flows. In J. A. Frankel (Ed.),
The Regionalization of the World Economy. Chicago and London: University of Chicago Press.
Felbermayr, G. J. and W. Kohler (2006). Exploring the intensive and extensive margins of world trade.
Review of World Economics (Weltwirtschaftliches Archiv) 127 (4), 642–674.
Frankel, J. A. and A. Rose (2000). Estimating the effect of currency unions on trade and output. National
Bureau of Economic Research Working Paper 7857.
Frankel, J. A., E. Stein, and S. J. Wei (1997). Regional trading blocs in the world economic system.
Wahington, D.C.: Institute for International Economics.
Helliwell, J. F. (1998). How much do national borders matter? Brookings institution (Washington, D. C.).
Leamer, E. E. and J. Levinsohn (1995). International trade theory: The evidence. In G. M. Grossmand and
K. Rogoff (Eds.), Handbook of Internatinal Economics, Volume 3. Elsevier.
Soloaga, I. and L. A. Winters (2001). Regionalism in the nineties: What effect on trade? Centre for
Economic Policy Research Working Paper 2183.
8

1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 0
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 1
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 2
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 3
Figure 1: The sequence of the estimated distance coefficients b
βt
5with t= 1970, ..., 2000 (solid line) using the basic gravity equation (1) and the fitted linear
trend (dashed line): All products combined and one-digit ISIC categories.
9

1980 2000
−2
−1
ISIC 31
1980 2000
−2
−1
ISIC 321
1980 2000
−2
−1
ISIC 322
1980 2000
−2
−1
ISIC 323
1980 2000
−2
−1
ISIC 324
1980 2000
−2
−1
ISIC 331
1980 2000
−2
−1
ISIC 332
1980 2000
−2
−1
ISIC 341
1980 2000
−2
−1
ISIC 342
1980 2000
−2
−1
ISIC 351
1980 2000
−2
−1
ISIC 352
1980 2000
−2
−1
ISIC 353
1980 2000
−2
−1
ISIC 355
1980 2000
−2
−1
ISIC 356
1980 2000
−2
−1
ISIC 361
1980 2000
−2
−1
ISIC 362
1980 2000
−2
−1
ISIC 369
1980 2000
−2
−1
ISIC 371
1980 2000
−2
−1
ISIC 372
1980 2000
−2
−1
ISIC 381
1980 2000
−2
−1
ISIC 382
1980 2000
−2
−1
ISIC 383
1980 2000
−2
−1
ISIC 384
1980 2000
−2
−1
ISIC 385
1980 2000
−2
−1
ISIC 390
Figure 2: The sequence of the estimated distance coefficients b
βt
5with t= 1970, ..., 2000 (solid line) using the basic gravity equation (1) and the fitted linear
trend (dashed line): Manufacturing industries broken down by three-digit ISIC categories.
10

1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 0
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 1
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 2
1970 1980 1990 2000
−1.5
−1.0
−0.5
ISIC 3
Figure 3: The sequence of the estimated distance coefficients b
βt
5with t= 1970, ..., 2000 (solid line) using the generalized gravity equation (2) and the fitted
linear trend (dashed line): All products combined and one-digit ISIC categories
11

1980 2000
−2
−1
ISIC 31
1980 2000
−2
−1
ISIC 321
1980 2000
−2
−1
ISIC 322
1980 2000
−2
−1
ISIC 323
1980 2000
−2
−1
ISIC 324
1980 2000
−2
−1
ISIC 331
1980 2000
−2
−1
ISIC 332
1980 2000
−2
−1
ISIC 341
1980 2000
−2
−1
ISIC 342
1980 2000
−2
−1
ISIC 351
1980 2000
−2
−1
ISIC 352
1980 2000
−2
−1
ISIC 353
1980 2000
−2
−1
ISIC 355
1980 2000
−2
−1
ISIC 356
1980 2000
−2
−1
ISIC 361
1980 2000
−2
−1
ISIC 362
1980 2000
−2
−1
ISIC 369
1980 2000
−2
−1
ISIC 371
1980 2000
−2
−1
ISIC 372
1980 2000
−2
−1
ISIC 381
1980 2000
−2
−1
ISIC 382
1980 2000
−2
−1
ISIC 383
1980 2000
−2
−1
ISIC 384
1980 2000
−2
−1
ISIC 385
1980 2000
−2
−1
ISIC 390
Figure 4: The sequence of the estimated distance coefficients b
βt
5with t= 1970, ..., 2000 (solid line) using the generalized gravity equation (2) and the fitted
linear trend (dashed line): Manufacturing industries broken down by three-digit ISIC categories
12

−0.50 −0.25 0.00 0.25 0.50
−0.25
0.00
0.25
0.50
^
ρ = 0.97
OLS2ave × OLS4ave
−0.50 −0.25 0.00 0.25 0.50
−0.25
0.00
0.25
0.50
^
ρ = 0.94
OLS2ave × OLS2aux
−0.50 −0.25 0.00 0.25 0.50
−0.25
0.00
0.25
0.50
^
ρ = 0.96
OLS4ave × OLS4aux
−0.50 −0.25 0.00 0.25 0.50
−0.25
0.00
0.25
0.50
^
ρ = 0.95
OLS2aux × OLS4aux
Figure 5: Cross plot of calculated relative change in the values of the estimated distance coefficient between
the beginning of the sample and of the end of the sample; based on Tables 1 and 2
13

1980 2000
0.1
0.2 im31
1980 2000
0.1
0.2 im321
1980 2000
0.1
0.2 im322
1980 2000
0.1
0.2 im323
1980 2000
0.1
0.2 im324
1980 2000
0.1
0.2 im331
1980 2000
0.1
0.2 im332
1980 2000
0.1
0.2 im341
1980 2000
0.1
0.2 im342
1980 2000
0.1
0.2 im351
1980 2000
0.1
0.2 im352
1980 2000
0.1
0.2 im353.4
1980 2000
0.1
0.2 im355
1980 2000
0.1
0.2 im356
1980 2000
0.1
0.2 im361
1980 2000
0.1
0.2 im362
1980 2000
0.1
0.2 im369
1980 2000
0.1
0.2 im371
1980 2000
0.1
0.2 im372
1980 2000
0.1
0.2 im381
1980 2000
0.1
0.2 im382
1980 2000
0.1
0.2 im383
1980 2000
0.1
0.2
im384
1980 2000
0.1
0.2 im385
1980 2000
0.1
0.2 im390
Figure 6: Share of each manufacturing industry in total trade of all manufacturing products among OECD countries
14

−0.5 0.0 0.5 1.0
−0.2 −0.1 0.0 0.1 0.2 0.3 0.4
Relative change in share in total manufacturing trade
Relative change in the value of distance coefficient
31
321
322
323
324
331
332
341
342
351
352
353−4
355
356
361
362
369
371 372
381
382
383
384
385
390
Figure 7: Cross plot of calculated relative change in the values of the estimated distance coefficient (OLS4ave)
and of relative change in share in total manufacturing trade between the beginning and of the end of the
sample period; based on Tables 2 and 3
15

−0.04 −0.02 0.00 0.02 0.04 0.06
−0.2 −0.1 0.0 0.1 0.2 0.3 0.4
Absolute change in share in total manufacturing trade
Relative change in the value of distance coefficient
31
321
322
323
324
331
332
341
342
351
352
353−4
355
356
361
362
369
371 372
381
382
383
384
385
390
Figure 8: Cross plot of calculated relative change in the values of the estimated distance coefficient (OLS4ave)
and of absolute change in share in total manufacturing trade between the beginning and of the end of the
sample period; based on Tables 2 and 3
16

Table 1: Estimated change in the absolute and relative values of the distance coefficient (auxiliary regression)
OLS2aux OLS4aux
ISIC Sector initial last absolute relative ISIC initial last absolute relative
change change change change
0 All products -0.654 -0.733 -0.079 -0.121 0 -0.668 -0.690 -0.023 -0.034
1 Agriculture -0.478 -0.515 -0.036 -0.076 1 -0.479 -0.588 -0.108 -0.226
2 Mining, quarrying -1.144 -1.670 -0.527 -0.461 2 -1.156 -1.686 -0.530 -0.458
3 Manufacturing -0.754 -0.721 0.033 0.043 3 -0.771 -0.670 0.101 0.131
31 Food, beverages, tobacco -0.352 -0.380 -0.029 -0.082 31 -0.369 -0.354 0.015 0.040
321 Textiles -0.526 -0.699 -0.173 -0.328 321 -0.544 -0.731 -0.187 -0.344
322 Wearing apparel -1.352 -1.113 0.239 0.177 322 -1.390 -1.275 0.115 0.083
323 Leather, leather products -0.910 -0.574 0.337 0.370 323 -0.940 -0.546 0.394 0.419
324 Footwear -1.183 -0.742 0.441 0.373 324 -1.224 -0.804 0.421 0.344
331 Wood, wood products -1.248 -1.201 0.046 0.037 331 -1.292 -1.106 0.185 0.144
332 Furniture -1.454 -1.178 0.276 0.190 332 -1.484 -1.152 0.332 0.224
341 Paper, paper products -1.689 -1.577 0.112 0.066 341 -1.721 -1.393 0.328 0.190
342 Printing, publishing -1.133 -0.932 0.201 0.177 342 -1.163 -0.859 0.304 0.262
351 Industrial chemicals -1.342 -1.068 0.274 0.204 351 -1.357 -0.963 0.394 0.290
352 Other chemical products -1.260 -0.879 0.382 0.303 352 -1.283 -0.751 0.532 0.415
353-4 Petroleum refineries and products -1.682 -1.874 -0.192 0.114 353 -1.716 -1.793 -0.076 -0.045
355 Rubber products -1.196 -0.913 0.284 0.237 355 -1.209 -0.911 0.298 0.246
356 Plastic products -1.088 -0.984 0.104 0.096 356 -1.122 -0.895 0.227 0.202
361 Pottery, china, earthware -0.846 -0.805 0.041 0.048 361 -0.866 -0.853 0.013 0.015
362 Glass, glass products -1.131 -1.061 0.070 0.062 362 -1.154 -1.089 0.066 0.057
369 Structural clay products -1.420 -1.066 0.354 0.250 369 -1.426 -1.082 0.345 0.242
371 Iron and steel basic industries -1.593 -1.744 -0.151 -0.095 371 -1.610 -1.661 -0.051 -0.032
372 Basic non-ferrous metals -1.361 -1.443 -0.083 -0.061 372 -1.368 -1.357 0.011 0.008
381 Fabricated metal products -1.101 -1.002 0.099 0.090 381 -1.127 -0.949 0.179 0.158
382 Machinery -1.137 -0.875 0.262 0.231 382 -1.159 -0.750 0.409 0.353
383 Electrical machinery -1.333 -0.826 0.507 0.380 383 -1.353 -0.739 0.614 0.454
384 Transport equipment -1.490 -0.978 0.513 0.344 384 -1.517 -0.892 0.625 0.412
385 Measuring, photo, optical equipment -0.916 -0.678 0.238 0.260 385 -0.943 -0.503 0.440 0.466
390 Other manufacturing -0.968 -0.753 0.215 0.222 390 -1.000 -0.703 0.298 0.298
17

Table 2: Estimated change in the absolute and relative values of the distance coefficient (average of the three first and three last values)
OLS2ave OLS4ave
ISIC Sector initial last absolute relative ISIC initial last absolute relative
change change change change
0 All products -0.647 -0.71 -0.062 -0.096 0 -0.657 -0.683 -0.026 -0.040
1 Agriculture -0.55 -0.569 -0.019 -0.034 1 -0.538 -0.611 -0.074 -0.137
2 Mining, quarrying -1.136 -1.631 -0.494 -0.435 2 -1.142 -1.626 -0.484 -0.424
3 Manufacturing -0.739 -0.684 0.055 0.074 3 -0.757 -0.655 0.102 0.135
31 Food, beverages, tobacco -0.452 -0.405 0.048 0.105 31 -0.460 -0.399 0.061 0.132
321 Textiles -0.531 -0.648 -0.116 -0.219 321 -0.527 -0.666 -0.138 -0.262
322 Wearing apparel -1.152 -1.062 0.089 0.078 322 -1.133 -1.159 -0.026 -0.023
323 Leather, leather products -0.968 -0.597 0.371 0.383 323 -0.970 -0.573 0.397 0.410
324 Footwear -0.915 -0.673 0.242 0.264 324 -0.901 -0.713 0.188 0.209
331 Wood, wood products -0.961 -0.891 0.069 0.072 331 -0.973 -0.832 0.141 0.145
332 Furniture -1.316 -1.076 0.24 0.182 332 -1.333 -1.080 0.254 0.190
341 Paper, paper products -1.395 -1.268 0.127 0.091 341 -1.445 -1.166 0.280 0.193
342 Printing, publishing -1.121 -0.87 0.251 0.224 342 -1.137 -0.837 0.299 0.263
351 Industrial chemicals -1.365 -1.031 0.334 0.244 351 -1.382 -0.958 0.424 0.307
352 Other chemical products -1.281 -0.83 0.451 0.352 352 -1.301 -0.752 0.549 0.422
353-4 Petroleum refineries and products -1.724 -1.922 -0.199 -0.115 353 -1.737 -1.812 -0.075 -0.043
355 Rubber products -1.202 -0.798 0.404 0.336 355 -1.214 -0.804 0.410 0.337
356 Plastic products -1.127 -0.881 0.246 0.218 356 -1.144 -0.847 0.298 0.260
361 Pottery, china, earthware -0.829 -0.822 0.007 0.008 361 -0.825 -0.865 -0.040 -0.049
362 Glass, glass products -1.147 -0.86 0.288 0.251 362 -1.156 -0.873 0.282 0.244
369 Structural clay products -1.384 -0.96 0.424 0.306 369 -1.398 -0.978 0.420 0.301
371 Iron and steel basic industries -1.363 -1.417 -0.054 -0.039 371 -1.393 -1.377 0.016 0.012
372 Basic non-ferrous metals -1.271 -1.316 -0.044 -0.035 372 -1.316 -1.288 0.028 0.021
381 Fabricated metal products -1.192 -0.934 0.258 0.216 381 -1.215 -0.910 0.305 0.251
382 Machinery -1.191 -0.857 0.335 0.281 382 -1.232 -0.793 0.439 0.356
383 Electrical machinery -1.254 -0.799 0.455 0.363 383 -1.286 -0.757 0.528 0.411
384 Transport equipment -1.376 -0.846 0.53 0.385 384 -1.409 -0.813 0.596 0.423
385 Measuring, photo, optical equipment -1.052 -0.69 0.362 0.344 385 -1.088 -0.604 0.484 0.445
390 Other manufacturing -0.931 -0.757 0.174 0.187 390 -0.938 -0.740 0.198 0.211
18

Table 3: Share of individual manufacturing industries in total manufacturing trade among
OECD countries
ISIC Sector 1970 2000 Absolute Relative
change change
31 Food, beverages, tobacco 0.0864 0.0615 -0.0249 -0.288
321 Textiles 0.0524 0.0245 -0.0280 -0.534
322 Wearing apparel 0.0227 0.0117 -0.0110 -0.483
323 Leather, leather products 0.0055 0.0033 -0.0022 -0.396
324 Footwear 0.0078 0.0051 -0.0028 -0.356
331 Wood, wood products 0.0202 0.0132 -0.0070 -0.347
332 Furniture 0.0047 0.0091 0.0044 0.935
341 Paper, paper products 0.0425 0.0307 -0.0118 -0.277
342 Printing, publishing 0.0099 0.0067 -0.0031 -0.317
351 Industrial chemicals 0.0735 0.0881 0.0146 0.198
352 Other chemical products 0.0271 0.0599 0.0328 1.210
353-4 Petroleum refineries and products 0.0246 0.0216 -0.0031 -0.124
355 Rubber products 0.0105 0.0103 -0.0001 -0.014
356 Plastic products 0.0078 0.0101 0.0022 0.284
361 Pottery, china, earthware 0.0036 0.0016 -0.0020 -0.566
362 Glass, glass products 0.0078 0.0057 -0.0021 -0.270
369 Structural clay products 0.0085 0.0068 -0.0018 -0.208
371 Iron and steel basic 0.0721 0.0191 -0.0530 -0.736
372 Basic non-ferrous metals 0.0473 0.0247 -0.0226 -0.478
381 Fabricated metal products 0.0350 0.0290 -0.0060 -0.171
382 Machinery 0.1510 0.1587 0.0077 0.051
383 Electrical machinery 0.0704 0.1245 0.0542 0.770
384 Transport equipment 0.1630 0.2161 0.0531 0.326
385 Measuring, photo, optical equipment 0.0299 0.0418 0.0119 0.400
390 Other manufacturing 0.0158 0.0164 0.0006 0.036
19