Download Productivity, resource endowment, and trade

Productivity, resource endowment, and trade
performance of the wood product sector
Bertrand M. Koebela , Anne-Laure Levetb , Phu Nguyen-Vana ,
Indradev Purohoob , and Ludovic Guinardb
a
BETA, CNRS and Université de Strasbourg
b
FCBA, Paris
March 6th, 2015
Abstract
This paper aims to analyse the determinants of international trade of wood
products, considering three main groups: woodworking products, pulp and paper
and wooden furniture. We extend the Heckscher-Ohlin-Vanek framework in order
to take into account industrial performance factors. Empirical tests are based on
data on European countries between 1995 and 2007. The results are consistent
with the HOV hypothesis which means that the forest resource endowment is an
important determinant for explaining differences in net trade of the three groups
of wood products. However, empirical tests also show limits of the HOV model
for explaining international trade of wood products. Indeed, factors reflecting
industrial performance of wood sectors, including total factor productivity and
average labour cost, have a significant role in determining differences in net trade
of wood products.
Keywords: Heckscher-Ohlin-Vanek hypothesis; trade; wood products
JEL Classification: F10; L60; Q23
1
1
Introduction
Considering the main exporters of wood products at worldwide, it appears that there
is no systematic correlation between the forest resource and the trade balance of wood
products (see Figure 1). At a comparable level of forest resource (measured in hectares
per inhabitant), some countries benefit from a positive trade balance of wood products
(Germany, Austria, Italy, etc.) whereas some other suffer from a deficit (France, Japan,
USA).
Figure 1: Trade balance and forest area. Data sources: UN Comtrade and FAO 2010.
An important part of the literature, however, explains international trade between
countries by differences in endowment of production factors according to the HeckscherOhlin model. More precisely, countries for which the required production factors are
relatively abundant have comparative advantages in the production of those goods,
and therefore export them (Heckscher 1919, Ohlin 1933). The multi-factor and multiproduct model has been developed by Vanek (1968) and is known under the HeckscherOhlin-Vanek (HOV) hypothesis. In the forest-based sector, HOV model predicts that a
country with relatively more abundant forest resource will also have larger net exports
of wood products other things being equal (Prestemon and Buongiorno, 1997).
In this context, this paper aims to analyse the determinants of international trade
of wood products in order to highlight the key drivers of net exports of wood products.
2
This question is of particularly high importance in France which combines a structural
trade deficit in wood products and one of the most abundant forest resources in Europe.
Compared to the existing literature (Bonnefoi and Buongiorno 1990, Lundmark 2010,
Prestemon and Buongiorno 1997, Uusivuori and Tervo 2002), our empirical analysis
considers the forest resource endowment for explaining international trade of wood
products, but also other factors with particular focus on the industrial performance
indicators. Furthermore, this paper will consider processed and finished wood products
(woodworking products, pulp, paper and paperboard, furniture) whereas, so far, the
literature was more focused on forest products (roundwood) and first processed products
(pulp, panel).
The paper is organized as follows. The next section gives an overview of the international trade in wood products as its evolution since 2000 based on UN Comtrade
data. The international competitiveness of the main export countries is also analysed
through market shares and net exports. Section 3 describes the data. Section 4 presents
an econometric model for assessing the Heckscher-Ohlin-Vanek (HOV) hypothesis. The
objective is to quantify the importance of industrial performance indicators in the explanation of the trade performance of export countries with a focus on the total factor
productivity (TFP). Section 5 discusses the results and draws some recommendations.
Section 6 concludes.
2
International trade of wood products
In this paper, wood products include all forest-based and wood products from the forest
(roundwood) to finished wood products. We divide into four main groups reflecting
their type of industrial transformation (see Table A1 in the Appendix). The world
exports of wood products accounted 442 billion dollars in 2011, which represents 2,3
percent of the total international trade of commodities. The trade in value is largely
dominated by processed wood products (with higher added value) whereas rough wood
is marginal. Pulp and paper products represent more than the half of total exports
(54% in 2011) followed by woodworking products (23% in 2011) and furniture (20%
in 2011). The share of harvested wood products was about 3% of total trade. These
shares are relatively stable since 2000.
The most important exporters of wood products are mainly traditional forest countries from North America and Europe (USA, Canada, Sweden, Finland) and more
and more emerging countries (China, Russia, Brazil). Between 2000 and 2011, China
became the most important exporter of wood products in value, in particular in the
wooden furniture sector, which reflects its dynamic position in export markets. During
this time period, exports of China in wood products have been multiplied by 8. In
developed countries, only Germany, Sweden and Austria won market shares between
3
2000 and 2011 (see Table 1). On the opposite, USA, Canada, Finland and France
lost market shares. Especially, exports of Canada decreased by 18% between 2000 and
2011, mainly because of the fall of the US housing market and the decrease of exports
to Japan.
Table 1: Export market shares (percent) in 2000 and 2011 - all wood products
Countries 2000 2011
China
2.3 12.0
Germany
8.2 11.8
USA
11.1 10.1
Canada
15.1
7.5
Sweden
5.2
5.5
Italy
4.9
4.9
Poland
1.6
4.0
Finland
5.1
4.0
France
4.1
3.7
Austria
2.7
3.3
Russia
1.7
2.6
Brazil
1.9
2.5
Data source: UN Comtrade.
In terms of net exports, only Japan, the USA, and France have a negative trade
balance in wood products due to a high level of imports, in particular in wooden furniture and woodworking products (Figure 2). Despite a decrease of its market share
in the last decade, Canada still has a significantly positive trade balance because of
relatively low imports. In comparison, China and Germany benefit from a less positive
trade balance because they also import a lot of wood products. Due to its relatively
low resource endowment, China has to import harvested products which provide the
raw material to wood processing industries (panel, furniture).
3
Data
The data were collected from several sources: UN Comtrade (for total net exports
of forest-based and wood products), Euklems (total factor productivity), World Bank
(GDP per capita), FAOSTAT (production and consumption of timber) and Eurostat
(for other variables). Three subsectors are considered: pulp and paper products, wooden
furniture, and woodworking products.1
1
Harvested products subsector is not included in the study because of data unavailability.
4
30
billion$
20
10
0
-10
-20
-30
Harvested products
Woodworking products
Pulp, paper and paperboard
Wooden furniture
Figure 2: Trade balance of main exporters. Data source: UN Comtrade 2011.
The data cover 15 European countries (Austria, Belgium, Czech Republic, Denmark,
Germany, Finland, France, Hungary, Ireland, Italy, the Netherlands, Slovenia, Spain,
Sweden, and the UK) for the period 1995-2007. The list of countries for which data are
available depend on the sub-sector considered. More precisely, for the wooden furniture
subsector 3 countries (Belgium, Italy, and Sweden) were missing from the above list
of 15 countries. The woodworking sub-sector includes all the 15 countries. For the
pulp and paper subsector, there are 14 countries, only Belgium being excluded from
the regressions.
These different sources allow us to construct a sample that includes all the variables
of interest. More details on variable definitions are given in Table A2 in the Appendix.
The descriptive statistics for these variables are reported in Table 2. Variables are
divided in the following groups:
• Trade performance. We use T vol as a measure of trade performance. It is defined
by exports minus imports.
• Income. We use gross domestic products (GDP ) to proxy the domestic demand
for wood products. We expect that an increase in income rises the demand for
5
wood products, leading to a deterioration of trade performance indicator for these
products.
• Resource endowment. These variables measure the relative advantage of a country compared to others. In terms of wood products, we use either production of
roundwood (P rodRW ) and net exports of roundwood (ExpRW ), or consumption
of roundwood (ConsRW ). We observe that P rodRW , ExpP RW , and ConsRW
cannot be used simultaneously in the same regression because they are highly
colinear. Due to its different position in the production chain, we retain production of wood-based panels (P rodW P ) and consumption of wood-based panels
(ConsW P ) as alternative measures for resource endowment for the wooden furniture subsector.
• Other factors. To control for other determinants of trade performance than those
highlighted by the HOV hypothesis, we also include other variables. Factors
thought to have an influence on trade performance are total factor productivity
(T F P ), average labour cost (LC), and average production value per firm (Y /N ).
4
A reappraisal of the Heckscher-Ohlin-Vanek hypothesis
According to the Heckscher-Ohlin-Vanek hypothesis, the country’s net exports of a
given good are an increasing function of its resource endowment and decreasing in its
income (Prestemon and Buongiorno 1997, Lundmark 2010). Several assumptions have
to be made for supporting the HOV model (Leamer 1984, Prestemon and Buongiorno
1997): (i) some factors are immobile between countries, (ii) markets are competitive
with no barriers to trade; (iii) the same technology is available to all producers; (iv)
consumption is increasing in income. These assumptions seem to be plausible when
one considers trade on wood products. Indeed, the resource endowment immobility is
true for most natural resources, in particular forests. Wood markets are reasonably
competitive in that products can move freely across countries (except for some forest
products which may have restrictions to export in some countries) and technologies are
widely available.
Regarding trade in forest products, the HOV hypothesis has been tested in few empirical studies leading to various results. In particular, Bonnefoi and Buongiorno (1990)
found empirical support to the HOV hypothesis: the forest endowment has a strong
positive effect on net trade for all considered commodities (roundwood, sawnwood,
wood-based panels, pulp and paper). Prestemon and Buongiorno (1997) tested the
same model for interstate trade in the United States and found a positive relationship
6
Table 2: Descriptive statistics
Variable
Obs.
Mean Std. Dev.
Min.
Max.
Pulp and paper products
T vol
156
295.32
5656.16 -8270.89 14881.94
GDP
156 72441.19 75110.45
33.14
292000
TFP
156 2687.58
2991.32
24
17834
LC
156
322.08
158.07
10
606
Y /N
156 1491.65
1754.03
4
7548
P rodRW
156
22.30
23.94
0.84
98.20
ConsRW
156
24.50
25.52
0.70
103.92
ExpRW
156 -2204.89
4099.98 -15458.26 3815.96
Wooden furniture
T vol
143
-67.32
535.89 -1945.24 1165.66
GDP
143 76737.24 77022.29
33.14
292000
TFP
143 2042.45
815.20
24
4285
LC
143
217.65
117.84
5
422
Y /N
143
96.85
76.96
2
327
P rodRW
143
18.65
20.69
0.84
76.73
ConsRW
143
20.50
21.77
0.70
74.19
ExpRW
143 -1846.08
4013.05 -15458.26 3815.96
P rodW P
143
3.19
3.71
0.01
17.71
ConsW P
143
3.27
3.41
0.30
14.84
Woodworking products
T vol
161
-531.63
3543.40 -9712.57 5226.04
GDP
161 70993.32 74326.70
33.14
292000
TFP
161 2139.70
759.99
22
4428
LC
161
237.89
115.98
7
465
Y /N
161
118.05
92.34
1
411
P rodRW
161
21.76
23.74
0.84
98.20
ConsRW
161
24.00
25.26
0.70
103.92
ExpRW
161 -2225.28
4032.94 -15458.26 3815.96
between the state’s net exports (lumber and wood products, paper and allied products)
and their forest endowment. Uusivuori and Tervo (2002) tested the Heckscher-OhlinVanek model for the trade of industrial roundwood and forest products in 18 OECD
countries. According to their results, the HOV hypothesis is partially confirmed. In
particular, they conclude that the role of resource is becoming less important in developed countries in shaping the development of forest industries because of other relevant
7
factors (technology, value added products). Finally, Lundmark (2010) explores which
factors (forest endowment, domestic demand, energy policies) could explain differences
in net trade of forest products between EU countries. The results provide mixed support to the HOV hypothesis in that the forest endowment appears as an important
determinant for explaining differences in net trade but not the domestic demand. In
addition, it is also important to consider other factors like renewable energy policies.
While existing empirical studies seek to explain the trade of harvested or semiprocessed products (sawnwood, panels), our paper considers finished wood products
(elements of construction, furniture, etc.). In addition, our work aims to go beyond
the HOV framework (which is limited to resource endowment and income) to include
other determinants of trade performance of wood products. Following Vanek (1963)
and later Lundmark (2010), we propose the following formulation to reappraisal the
Heckscher-Ohlin-Vanek hypothesis:
T rade perf ormanceit = β0 + β1 Incomeit + β2 Resource endowmentit
+γOther f actorsit + λt + µi + εit
(1)
where T rade perf ormance is exports minus imports (T vol), Income is a measure of
income, e.g. gross domestic products (GDP ), Resource endowment is either production of roundwood (P rodRW ), net exports of roundwood (ExpRW ) or consumption
of roundwood (ConsRW ). In the case of wooden furniture subsector, we also use
alternative measures of resource endowment such as production of wood-based panels (P rodW P ) and consumption of wood-based panels (ConsW P ). Other f actors
includes total factor productivity (T F P ), average labour costs (LC), and average production value per firm (Y /N ). Time effects λt are included via year dummies in order
to account for the time effects that are common to all countries. Individual effects µi
help to control for unobserved heterogeneity specific to countries.
According to the HOV theory, the endowment variable should have a positive linear effect on net trade of wood products (i.e. positive effects of P rodRW , ConsRW ,
P rodW P , ConsW P ). As an increase in ExpRW represents a reduction of resource
endowment, its effect of trade should have an expected negative sign. Moreover, following the HOV hypothesis, the income variable should have a negative effect on trade
performance. In addition, for other factors, we expect a positive relationship between
total factor productivity (T F P ) and the net trade as well as for the average firm size
(Y /N ). On the contrary, we expect to find a negative effect of average labour costs
(LC) on the trade performance variable.
5
Results
Tables 3-5 present results based on the specification in equation (1) for three groups of
wood products: pulp and paper, wooden furniture and woodworking products. Model
8
1 corresponds to the regressions which include production of roundwood P rodRW and
net exports of roundwood ExpRW as measures of resource endowment. Model 2 uses
consumption of roundwood ConsRW . Due to the position of the wooden furniture
subsector in the production chain, two alternative measures of resource endowment are
employed, production of wood-based panels P rodW P and consumption of wood-based
panels ConsW P .
Table 3: Estimation results of the HOV model,
Model 1
Variable
Coefficient
(Std.Err.)
GDP
-0.0134**
(0.005)
TFP
-0.024
(0.052)
LC
0.133
(0.462)
Y /N
-0.044
(0.090)
P rodRW
74.216**
(11.809)
ExpRW
-0.156**
(0.045)
ConsRW
–
Constant
Adjusted R2
Number of countries
Number of obs.
Hausman test
Endog. χ2 test
-962.196*
(500.947)
0.248
14
156
104.12**
0.017
pulp and paper products
Model 2
Coefficient
(Std.Err.)
-0.016**
(0.005)
-0.019
(0.052)
0.041
(0.464)
-0.034
(0.090)
–
–
77.087**
(11.804)
-826.675*
(499.599)
0.235
14
156
68.79**
0.037
Notes: Dependent variable: T vol. Regression includes country and time dummies (not reported here). Results correspond to the fixed effects estimator. An endogeneity χ2 test was
performed for T F P .
∗
and
∗∗
denote significance at the 10% and 5% levels, respectively.
For each model, we compute the Hausman test to compare two alternative estimators corresponding to fixed effects and random effects. The test reports that fixed
effects specification is always preferred to the random effects specification. Further9
Table 4: Estimation results of the HOV model, wooden furniture
Model 1a
Model 2a
Model 1b
Model 2b
Variable
Coefficient Coefficient Coefficient Coefficient
(Std.Err.) (Std.Err.) (Std.Err.) (Std.Err.)
GDP
-0.011**
-0.011**
-0.016**
-0.014**
(0.001)
(0.001)
(0.001)
(0.002)
TFP
0.010
0.013
0.018
0.020
(0.020)
(0.019)
(0.018)
(0.020)
LC
0.426**
0.425**
0.293*
0.172
(0.192)
(0.192)
(0.176)
(0.201)
Y /N
-0.763
-0.741
-0.457
-1.117*
(0.569)
(0.567)
(0.529)
(0.580)
P rodRW
20.518**
–
–
–
(4.736)
ConsRW
–
19.376**
–
–
(4.471)
ExpRW
-0.008
–
–
–
(0.016)
P rodW P
–
–
203.425**
–
(31.780)
ConsW P
–
–
–
168.299**
(49.180)
Constant
194.268
170.720
308.160** 322.129**
(150.513)
(146.841)
(117.422)
(138.623)
Adjusted R2
0.427
0.430
0.512
0.397
Number of obs.
143
143
143
143
Number of countries
13
13
13
13
Hausman test
109.08**
113.30**
115.18**
112.40**
2
Endog. χ test
1.281
1.549
2.797
2.214
Notes: Dependent variable: T vol. Regression includes country and time dummies (not reported here). Results correspond to the fixed effects estimator. An endogeneity χ2 test was
performed for T F P .
∗
and
∗∗
denote significance at the 10% and 5% levels, respectively.
more, the T F P variable, as constructed in the database Euklems, may be correlated
with unobserved factors included in the regression residuals. A test for the endogeneity
of T F P is performed. The result shows that T F P can be treated as an exogenous
regressor for all the regressions considered.
The estimation results show that GDP has the expected negative effect on trade
10
Table 5: Estimation results of the HOV model, woodworking products
Model 1
Model 2
Variable
Coefficient Coefficient
(Std.Err.)
(Std.Err.)
GDP
0.013*
0.013*
(0.007)
(0.007)
TFP
0.265**
0.271**
(0.103)
(0.102)
LC
0.234
0.322
(1.003)
(0.987)
Y /N
0.400
0.375
(1.458)
(1.454)
P rodRW
73.222**
–
(18.274)
ExpRW
-0.034
–
(0.071)
ConsRW
–
71.865**
(18.059)
Constant
-3065.485** -3159.541**
(745.345)
(723.640)
2
Adjusted R
0.044
0.049
Number of countries
15
15
Number of obs.
161
161
Hausman test
117.84**
100.31**
Endogeneity χ2 test
1.116
1.060
Notes: Dependent variable: T vol. Regression includes country and time dummies (not reported here). Results correspond to the fixed effects estimator. An endogeneity χ2 test was
performed for TFP.
∗
and
∗∗
denote significance at the 10% and 5% levels, respectively.
performance, except for the woodworking products. The effects of resource endowment,
P rodRW , ConsRW , P rodW P , and ConsW P have the expected positive signs. Net
exports of roundwood ExpRW has a significant effect for pulp and paper subsector
only and in this case, the effect of ExpRW has the expected negative sign. In sum, the
HOV hypothesis is globally verified for the data on the wood product sector.
Regarding other factors included in regressions, T F P has a significant and positive
effect on trade performance for woodworking products whereas the effect of LC is
significant and positive for the wooden furniture. While the sign of the effect of T F P
is as expected, that of LC seems counterintuitive. However, a high value of LC may
11
depict the situation that wooden furniture products are of high quality and that their
production need skilled labour (hence correlated with high personnel costs). Hence,
the positive effect of LC represent the phenomenon of increased exports of high quality
products in the wooden furniture subsector. The effect of Y /N , negative and significant
only in one specification for wooden furniture sector, appears unexplainable. Moreover,
none of T F P , LC, and Y /N has a significant effect when the pulp and paper products
are considered.
In summary, empirical tests conducted in this study show limits to the use of HOV
model for explaining wood products international trade. Firstly, in some cases, the
roundwood consumption, which includes imports of roundwood from abroad, shows
the same statistical significance as the roundwood production. In that case, the endowment variable reflects more raw material supply for forest industries than local forest
resources. In addition, in the furniture sector, the panel production, which is a semiprocessed wood product, is more significant than the roundwood production. This can
be explained by the growing share of furniture produced from panels (particle board,
medium-density fibreboard) and the less importance of furniture produced from sawnwood. But it also suggests that the more the wood product is transformed, the less its
production depends directly on the local resource endowment. Secondly, the empirical
results highlight the role of industrial performance through TFP, which is a significant
determinant for explaining differences in net trade of woodworking products. Thirdly,
the overall fit of the model (measured by adjusted R2 ) is relatively low according to the
groups of wood products (from 5% to 50%). This suggests that only a (small) part of
the determinant factors has been captured in the model, which needs to explore other
factors, in particular beyond the resource endowment. These results are somewhat
different from Prestemon and Buongiorno (1997) and Lundmark (2010) who found a
higher adjusted R2 (from 35% to 90%). One explanation could be that these previous
studies focus on semi-transformed products whereas the current paper includes mainly
finished wood products. Finally, the heterogeneity of the results between the three
main groups of wood products (pulp and paper, furniture and woodworking products)
underlines the need to adopt a specific approach (and thus modelling) for each of these
groups.
6
Conclusion
The purpose of this paper was to test the Heckscher-Ohlin-Vanek theory in international forest-based and trade in wood products, including finished wood products. In
particular, the objective was to analyse the role of the forest resource endowment for
explaining wood products trade as well as, beyond the resource endowment, to explore
the role of other determinants reflecting the industrial performance of sectors. For this
12
purpose, an extended specification of the HOV model has been defined and empirically
tested for European countries and three groups of wood products.
The results show that the forest resource endowment, measured by the roundwood
production, is an important determinant for explaining differences in net trade of the
three groups of wood products. The domestic demand, measured by the GDP, is also
determinant for pulp and paper products and wooden furniture but not for woodworking products. These results are consistent with the HOV predictions. However,
empirical tests conducted in this study also show limits to the use of HOV model for
explaining wood products international trade. It appears that the endowment variable
is not enough for explaining trade patterns of wood (finished) products and underlines
the needs for exploring other factors. In particular, industrial performance indicators,
including total factor productivity and average labour costs, may have an influence on
determining trade performance of wood products sectors even though their effects depend on the type of subsector considered. However, in several cases they are not found
to be significant. This may be due to the wrong level of aggregation here and the invalid
hypothesis of a representative firm. If within a country, high value added and high wage
firms coexist with small firms producing standardized products at a lower wage, then
an aggregate analysis cannot identify the relationships of interest. The role of industrial
performance could be then more deeply analysed by using other trade models based
on firm heterogeneity (e.g. Melitz model) and firm data. A complementary study of
export behavior in the wood sectors based on firm data is proposed by Koebel et al.
(2015).
7
Acknowledgement
This research was financially supported by the GIP ECOFOR.
13
Appendix
Table A1: List of wood products
Harvested products
Roundwood, sawlogs,
wood fuel, ...
Other forest products
Woodworking products
Sawnwood
Veneer sheets
& wood-based panels
Packaging
Parquets
Carpentry and joinery
Miscellaneous
Pulp and paper
Pulp
Paper & paperboard
Wooden furniture
Kitchen furniture
Furniture for dining-room,
living-room, ...
Office furniture
Table A2: Definition of variables
Variable name
T vol
P rodRW
ExpRW
ConsRW
P rodW P
ConsW P
GDP
TFP
LC
Y /N
Definition
Trade performance (exports minus imports) in kg
Production of roundwood, in million m3
Net exports of roundwood, in million m3
Apparent consumption of roundwood, in million m3
Production of wood-based panels, in million m3
Apparent consumption of wood-based panels, in million m3
Gross domestic product in PPP, in billion dollars
Total factor productivity (index)
Average labour cost, in euros/person
Average production value per firm, in euros/firm
14
Nature
UN COMTRADE
FAOSTAT
FAOSTAT
FAOSTAT
FAOSTAT
FAOSTAT
World Bank
EU Klems
Eurostat
Eurostat
References
Koebel B.M., Levet A.-L., Nguyen-Van P., Olland F. (2015), “Fixed costs and total
factor productivity in the French wood sector,” unpublished manuscript.
Bonnefoi B., Buongiorno J. (1990), “Comparative advantage of countries in forestproducts trade,” Forest Ecology and Management 36, 1-17.
Heckscher E. (1919), “The effects of foreign trade on the distribution of income,”
Ekonomisk Tidskriff 21, 1-32.
Leamer E.E. (1984), Sources of International Comparative Advantage, Boston, MA:
MIT Press.
Lundmark R. (2010), “European trade in forest products and fuels,” Journal of Forest
Economics 16, 235-251.
Ohlin B. (1933), Interregional and International Trade, MA: Harvard University Press.
Prestemon J.P., Buongiorno J. (1997), “Comparative advantage in US interstate forest
products trade,” Journal of Forest Economics 3, 207-228.
Uusivuori J., Tervo M. (2002), “Comparative advantage and forest endowment in forest
products trade: Evidence from panel data of OECD countries”, Journal of Forest
Economics 8, 53-75.
Vanek J. (1963), The Natural Resources Content of United States Foreign Trade, 18701955. MIT Press, Massachusetts.
Vanek J. (1968), “The factor proportions theory: The n factor case,” Kyklos 21, 749-756.
15