U n i t e d n at i o n s C o n f e r e n C e o n t r a d e a n d d e v e l o p m e n t
POLICY SPACE IN AGRICULTURAL MARKETS
POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIESRESEARCH STUDY SERIES No. 73
Printed at United Nations, Geneva1600845 (E) – January 2016 – 245
UNCTAD/ITCD/TAB/75
United Nations publicationISSN 1607-8291
POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
RESEARCH STUDY SERIES No. 73
POLICY SPACE IN AGRICULTURAL MARKETS
by
Alain McLaren
UNCTAD, Geneva
New York and Geneva, 2016
U N I T E D N AT I O N S C O N F E R E N C E O N T R A D E A N D D E V E L O P M E N T
ii POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
Note
The purpose of studies under the Research Study Series is to analyse policy issues and to
stimulate discussions in the area of international trade and development. The Series includes
studies by UNCTAD staff and by distinguished researchers from other organizations and academia.
The opinions expressed in this research study are those of the authors and are not to be
taken as the official views of the UNCTAD secretariat or its member States. The studies published
under the Research Study Series are read anonymously by at least one referee. Comments by
referees are taken into account before the publication of studies.
The designations employed and the presentation of the material do not imply the
expression of any opinion on the part of the United Nations concerning the legal status of any
country, territory, city or area, or of authorities or concerning the delimitation of its frontiers or
boundaries.
Comments on this paper are invited and may be addressed to the author, c/o the
Publications Assistant, Trade Analysis Branch (TAB), Division on International Trade in Goods and
Services, and Commodities (DITC), United Nations Conference on Trade and Development
(UNCTAD), Palais des Nations, CH-1211 Geneva 10, Switzerland; e-mail: [email protected]; fax no:
+41 22 917 0044. Copies of studies under the Research Study Series may also be obtained from
this address.
Studies under the Research Study Series are available on the UNCTAD website at
http://unctad.org/tab.
Series Editor:
Chief Trade Analysis Branch
DITC/UNCTAD
UNCTAD/ITCD/TAB/75
UNITED NATIONS PUBLICATION
ISSN 1607-8291
© Copyright United Nations 2016 All rights reserved
Policy Space in Agricultural Markets iii
Abstract
As an outcome of the Uruguay Round Agreement on Agriculture, all agricultural products now have a
bound tariff rate on their imports. This system of bound tariffs combines the rigidity of an upper limit
that is independent of future economic conditions but discretion as governments have a whole array of
choices in terms of applied tariffs as long as they are set below the bound rate. One recurring
argument is that bound rates may limit countries’ policy flexibility, or policy space, in response to
particular economic circumstances. This paper looks at the use and availability of this policy space in
agricultural markets. This is first done in a descriptive setting, then by assessing what plays a role in
determining this space using an empirical analysis. A general finding is that policy space in agricultural
products is generally available, and only limited for developed countries. Many developing countries
have ample room to raise tariffs in most agricultural imports without infringing binding commitments.
For LDCs there is virtually no imports for which policy space is not available. The findings indicate that
four specific factors are related to the use of policy space, which are the elasticity of import demand,
the fact that the goods are being used as intermediates, food security and protection of local
producers. The results suggest that policy space tends to be used relatively less for products with
lower elasticity of demand and for intermediate products. In regard of products relevant for food
security, the results suggest that policy space is larger. In regard to products that face domestic
competitors, the results indicate lower tariff water and more use of policy space, suggesting that
producer protection is an issue related to the level of policy space to use and the level of market
protection to set.
Keywords: International trade, policy space, WTO, tariffs.
JEL Classification: F10, F13
iv POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
Acknowledgements
This study was prepared during my stay in the Trade Analysis Branch under the Division for International Trade and Commodities at UNCTAD. The author wishes to thank Marco Fugazza,
Alessandro Nicita, and Ralf Peters for discussion and comments. This paper represents the personal views of the author only, and not the views of the UNCTAD
secretariat or its member States. The author accepts sole responsibility for any errors remaining.
Policy Space in Agricultural Markets v
Contents
1 INTRODUCTION .......................................................................................................................... 1
2 DATA ........................................................................................................................................... 2
3 POLICY SPACE ........................................................................................................................... 3
4 EMPIRICAL SPECIFICATION ................................................................................................... 10
5 ROBUSTNESS CHECKS ........................................................................................................... 15
6 CONCLUSION ........................................................................................................................... 17
BIBLIOGRAPHY ...................................................................................................................................... 18
vi POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
List of figures
Figure 1. Policy space and the share of trade covered ........................................................................ 4
Figure 2. Policy space and gross domestic product per capita ........................................................... 5
Figure 3. Policy space in least developed countries, at the HS 2 digit level ........................................ 6
Figure 4. Policy space in developing non-LDCs, at the HS 2 digit level .............................................. 7
Figure 5. Share of imports from partners within RTAs, by income level .............................................. 8
Figure 6. Policy space .......................................................................................................................... 9
Figure 7. Policy space for imported and produced goods ................................................................... 9
List of tables
Table 1. OLS and fractional heteroskedastic probit .......................................................................... 12
Table 2. Water and true water by income group (OLS) ..................................................................... 13
Table 3. MFN/bound and true MFN/bound by income group (OLS) ................................................. 14
Table 4. MFN/bound and true MFN/bound by income group (Fract. Heter. Probit) ......................... 14
Table 5. OLS and fractional heteroskedastic probit at the HS 6 digit level ...................................... 15
Table 6. OLS and fractional heteroskedastic probit excluding the import demand elasticity .......... 16
Policy Space in Agricultural Markets 1
1. INTRODUCTION
Trade in agri-food products is governed by a set of rules influencing market access conditions by
limiting the level of protection countries can legitimately apply. At the multilateral level, the rules
governing the trade of agricultural products are those agreed in the Uruguay Round Agreement on
Agriculture (URAA). The URAA was concluded in 1994 and became fully implemented by 2005. As an
outcome of the URAA each WTO member replaced border barriers with an equivalent tariff. This tariff is
known as the "bound rate" and refers to the highest rate the country could then apply without infringing
the agreement.1 The rationale behind setting tariff ceilings was to increase policy certainty and limit
negative spillovers of domestic policies on international markets. As noted in Horn et al. (2010), the
system of bound tariffs combines the rigidity of an upper limit that is independent of future economic
conditions but discretion as governments have a whole array of choices in terms of applied tariffs as
long as it stays below the bound rate. Still, one recurring argument is that bound rates may limit
countries’ policy flexibility (or policy space) in response to particular economic circumstances.
In reality, countries often choose to apply a tariff that is well below the bound rate, and thus generally
maintain significant flexibility in raising the tariffs of many agricultural products. This flexibility can be
measured by the difference between the bound and the “most favoured nation” tariff rate (MFN) and is
referred to as tariff water or binding overhang. In order to understand the use that developing countries
do, or don't do, of such flexibility, this paper examines the availability and use of policy space related to
agricultural trade.2
There are a certain number of reasons for which the desired MFN is often set below the bound rate and
thus policy space remains available. Theoretical work by Amador and Bagwell (2012) considers that
governments set tariffs so as to maximize a weighted average of consumer surplus, tariff revenue and
profits in the import-competing sector. They will be influenced by many economic factors, some of
which will affect several of them simultaneously. This paper investigates some of the most relevant
factors that can influence the use of policy space. In particular, the paper examines the relationship
between policy space and the elasticity of import demand, the fact that the goods are being used as
intermediates, food security and protection of local producers.
The elasticity of import demand will have a direct impact on consumer surplus and will influence the
amount of tariff revenue through changes in import volumes. For example, if a government's objective
is to reduce demand of imports, small changes in tariffs on products with elastic demand would be
rather effective, while any raise in tariff would not substantially impact imports of products with inelastic
demand which lack of domestic substitutes. In contrast, tariff setting would be quite different if the
government’s objective is to increase tariff revenues. In practice, governments under fiscal constraints
may be willing to use their policy space on products with inelastic demand. The use of policy space
may also be related to whether the good is used as an intermediate product. In such cases
policymakers may be inclined to keep the MFN applied relatively low thus favoring local processing
1 The guidelines were the following for developed countries, as found in Multilateral Trade Negotiations on Agriculture (2000): for previously bound tariff lines, they had to keep on using the same rate if there was no NTB, and if there was an
NTB they had to eliminate it or use the following tariffication formula: T = �������� ∗ 100.For lines that were not previously
bound, they had to use the rate that was applied as of September 1986 if there was no NTB, and if there was an NTB they had to use the tariffication formula. For developing countries, the guidelines were the same for lines that were already bound. For those that were not previously bound, they had the choice to do as developed countries or offer a ceiling binding.
2 The present study does not look at whether the bound rate has a useful purpose when it is well above the applied rate. The literature tends to agree on the fact that the bound rate may nonetheless play a positive role for several reasons. For instance, Bacchetta and Piermartini (2011) note that this situation implies increased tariff stability as well as reduced uncertainty that exporters face in terms of trade policy. They also mention that theoretical work by Francois and Martin (2004) showed that there are welfare gains from both the reduced variability of tariffs and their lower average level. Bacchetta and Piermartini (2011) also put forward the argument found in Sala et al. (2010), who argue that a bound rate above the applied rate may not directly affect the intensive margin that will be influenced by the applied rate but may give a signal to exporters wishing to enter that there is more stability on the market. This was confirmed empirically in Handley (2014).
2 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
industries. However, it is possible that government strategies aimed to integrate vertical production
processes use all their policy space so as to facilitate the emergence of domestic suppliers.3 A third
factor affecting the use of policy space is food security. In this regard, the main argument is to support
local production so as to guarantee food supply from domestic producers while minimizing the effect of
external shocks. However, countries with insufficient agricultural resources rely more on food supply
from international markets and therefore may be seeking food security by facilitating trade on foodstuff
thus making little use of any available policy space. Finally, a reason for which one may expect
governments to set higher applied tariffs could be related to the presence of producers’ lobbies which
influence government decisions related to trade policy. If there aren't any domestic producers then
there won't necessarily be an incentive to increase the price of the imported product.
In investigating the use of policy space this paper will first show some descriptive statistics of policy
space in agriculture, by using a snapshot of recent data, namely 2013. It shows the distribution of
policy space with respect to trade covered, the relationship between policy space and GDP per capita,
and finally the distribution of policy space by product and by income level groups. A first finding
indicates that in the vast majority of cases policy space is available, meaning that countries could raise
tariffs if they wanted to. The second step is to perform econometric analysis to shed light on what
determines policy space. This is done over a longer time span in order to benefit from the advantages
of panel data techniques. The results suggest that the elasticity of import demand, intermediate goods,
food security issues and protection of local producers are all correlated to the amount of policy space
that is available. The information on what plays a strong role in influencing policy space and in which
way can have important policy implications, especially in present times where negotiations on new
bound rates are having trouble reaching consensus, and in particular in agriculture which has also been
one of the tough areas in terms of trade policy negotiations.
The rest of the paper is organized as follows. Section 2 presents the data that will be used for the
descriptive statistics and regressions, section 3 shows and discusses some stylized facts concerning
policy space, section 4 presents the empirical specifications used, section 5 presents some robustness
checks and section 6 concludes.
2. DATA
The tariff data used comes from the UNCTAD TRAINS database and includes ad valorem equivalents
using the UNCTAD method 1. It includes bound rates and MFN applied rates as an import value
weighted average at the HS 6 digit level. The import data comes from the UN COMTRADE database,
food production data comes from the Food and Agricultural Organization (FAO) of the United Nations
and the elasticity of import demand from Kee et al. (2008). All data except for production covers the
period 2008-2013 at the HS 6 digit level, covering 98 importing countries. The production data stops in
2012 and is converted from the FAO classification to an equivalent HS 2 digit level.
3 This relates to international trade and value chains, notably discussed in reports such as Humphrey and Memedovic
(2006) and in the Organisation for Economic Co-operation and Development's (OECD) Mapping Global Value Chains (2012).
Policy Space in Agricultural Markets 3
3. POLICY SPACE
A first look at the data shows us that there is a reasonable amount of variation of policy space over
time, both increasing and decreasing. It is the changes in MFN applied that reflect the changes in
observed policy space.4 MFN applied tariff decreases were observed to be more frequent than tariff
increases, with about 9 percent of total lines (at the HS 6 digit level) that saw a decrease from one year
to the next against less than 6 percent that went up.5 However, the percentage of lines in the
agricultural data over the period 2008-2013 that increased and hit the bound was only of 0.2 percent of
all lines (and about 4 percent of increased tariffs), and just under 0.5 percent of all lines increased and
were over 80 percent of the bound after the increase (whether they already were above this threshold
or not). This shows us that although there is a reasonable amount of changes in the MFN applied rates,
both up and down, all of the available policy space is nearly never used up.
The space between applied and bound tariffs will be measured in two ways, once as a ratio (MFN
applied / bound) and once as a difference (bound - MFN applied). The first will tell us the percentage of
the available space that is used up whereas the second will give information on the specific amount
that can still be used. Two additional measures will be used which slightly modify the above mentioned
computations so as to account for the presence of prohibitive tariffs when the latter are below the
bound rate and for the presence of preferential tariffs which render the MFN applied tariff irrelevant.
These two measures are aggregated as in Kee et al. (2009) and in Foletti et al. (2011) in order to be at
the same level as the production data, as shown in equations (1) and (2).
� ����� ������,� =∑ ����,�,�∗ ���,�,�∑ ����,�,�∗ ���,�,��∋��" ($���,%,� − ����,%,�)%∋��( (1)
)* �+ ,+�-. /����,� =∑ ����,�,�∗ ���,�,�∑ ����,�,�∗ ���,�,��∋��" (0���,�,�1���,�,�)%∋��( (2)
Where 2���,%,� is the elasticity of import demand of product i within a given HS 2 digit product in country c, 3���,%,� is the total imports of that country in that product, $���,%,� is the bound rate and ����,%,� the MFN applied rate. A value for tariff water of 10 implies that the MFN tariff can be increased by 10
percentage points (i.e. from 10 to 20 percent ). Available policy space as shown in equation 2 can be
seen in terms of its use. This index indicates how much of the available policy space is used and varies
from 0 to 1. A value of 0.2 implies that the country is using only 20% of the policy space available in
that specific product. A value of 0 implies an MFN rate of 0 and a non-zero bound (so no policy space
is used). A value of 1 implies MFN=bound, thus all policy space is used. In most figures and tables this
is labeled as “MFN/bound”.
4 The products for which there was room for negotiation may have been influenced by some of the variables that influence the MFN applied rates and that are discussed below. This may also have been the case for some bound levels in previous rounds. For this study one must bear in mind that time invariant variables could have influenced some of the bound rates but other than that policy maker's choices in recent years have only been affecting the MFN applied rates.
5 This is consistent with findings such as those in Esteovadeordal et al. (2008), where they observe increases and decreases of MFN applied rates even when a bound rate has been set (in this case data isn't restricted to agricultural products). Studies such as Bacchetta and Piermartini (2011) find that the tendency for applied tariffs to increase is smaller in the presence of a bound rate compared to when there is none and the tendency for them to decrease is more likely in the presence of a bound.
4 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
0.2
.4.6
.81
0 50 100 150 200water
LDCs DevelopingDeveloped
Distribution of tariff water in termsof the value of trade
0.2
.4.6
.81
0 .2 .4 .6 .8 1pp
LDCs DevelopingDeveloped
Distribution of mfn / bound in termsof the value of trade
Availability and use of policy space is different across countries depending on the level of
development.6 Figure 1 shows us the amount of trade that is taking place under different levels of
policy space, as defined by tariff water and MFN/bound.
Figure 1
Policy Space and the Share of Trade Covered
(a) (b)
On one hand one can see that for LDCs there is virtually no imports for which policy space is not
available. On the other hand, developed countries' imports have no available policy space. This can
either be due to a bound of zero, which is the case for close to a half of HS 6 digit lines in developed
countries, or because the countries are setting their MFN rate as high as possible, leaving themselves
with no policy space. Differences in the availability and use of policy space are also evident when
comparing across different levels of GDP per capita, as can be seen in Figure 2.
6 Countries are categorized by geographic region as defined by the UN classification (UNSD M49). Developed countries comprise those commonly categorized as such in UN statistics. For the purpose of this paper transition economies, when not treated as a single group, are included in the broad aggregate of developing countries.
Policy Space in Agricultural Markets 5
05
1015
20m
fn /
boun
d
0 10000 20000 30000 40000GDP per capita
policy_space Fitted values
MFN / bound and GDP per capita
Figure 2
Policy Space and Gross Domestic Product Per Capita
(a) (b)
It illustrates the relationship between policy space and GDP per capita for countries with a GDP per
capita below 40'000 USD. The aggregation by country is done by weighting each product by the
imported value. More developed countries tend to have less policy space, whatever the indicator used.
This is a reflection of developed countries' lower bound rates rather than higher MFN rates.
Figures 3 and 4 plot availability and use of policy space vis-à-vis the importance of the good for the set
of developing countries differentiated by LDCs and non LDCs. The importance of each product
(horizontal axis) is measured by its weight in the import basket (imports of the product / total imports).
For example, a value of 0.05 implies that the given product represents 5 percent of the import basket.
This standardization allows for comparison of countries with different levels of imports. Only products
which represent more than 1 percent of total imports are taken into account. The right hand side of
Figures 3 and 4 also display the importance of the trade flow in terms of its actual trade value.
050
100
150
200
Tar
iff w
ater
0 10000 20000 30000 40000GDP per capita
water Fitted values
Tariff water and GDP per capita
6 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
050
100
150
200
Tar
iff w
ater
0 .1 .2 .3 .4 .5Trade importance
Policy space and trade importance
0.2
.4.6
.8m
fn /
boun
d
0 .1 .2 .3 .4 .5Trade importance
Policy space used and trade importance
Figure 3
Policy Space in Least Developed Countries, at the HS 2 Digit Level
(a) (b)
(c) (d)
The availability of policy space in LDCs as measured by tariff water is illustrated in Figure 3a. Products
that represent a very large share of imports still have a substantial amount of tariff water. Products with
the highest amounts of tariff water (top left corner of Figures 3a and 3b) tend to be those that do not
represent a too large share of a country's imports. The use of policy space in relation to the importance
of the product is illustrated in Figure 3c. LDCs' use of policy space is not correlated to trade
importance. With the exception of very few products on the top left corner of Figure 3c the amount of
policy space used is generally below 50%, and rarely surpasses about 30% for the most imported
products. Interestingly, almost all very strongly imported products are in cereals (10) or oils/fats (15),
but even for these products which may be of importance for food security the use of policy space is
relatively low. Overall, availability and use of policy space of LDCs (both when defined by tariff water or
by MFN / bound) appear to not be very correlated to the importance of the product. As can be seen in
Figure 4, the situation in developing countries is relatively similar to the one in LDCs with large amounts
of available policy space and limited use of policy space even in the most important products.
07-MWI
11-UGA
01-LSO
23-MLI08-RWA
20-GNB
24-BFA
02-SEN
12-UGA
20-BEN
22-NER
12-SEN07-NER
02-ZMB
04-TZA
24-UGA
20-NPL
08-LSO
08-SEN
04-NPL
07-SEN18-NPL
17-MLI
01-ZMB
21-TZA
11-SEN
12-RWA
22-MWI19-TZA
20-NER
09-NER
18-ZMB
20-MWI
07-MLI11-NPL
19-MWI
20-SEN11-BEN
09-ZMB
22-SEN
21-BEN22-BEN
11-MWI
17-NPL07-GNB
17-BFA
01-NPL
21-MDG
09-LSO
21-GNB
17-MWI11-TZA07-ZMB
24-RWA20-MLI
24-ZMB
02-GNB
20-RWA
23-LSO
04-MDG
17-ZMB
22-NPL07-RWA15-MLI19-RWA
20-BFA
11-ZMB
22-MLI
12-ZMB
20-LSO
24-TZA
19-UGA
16-LSO
08-ZMB
20-TGO
09-SEN
23-ZMB21-MWI
10-ZMB
04-GNB
21-UGA
22-MDG
24-NPL
17-LSO
17-BEN
04-MWI
24-MLI
21-LSO
11-MLI
19-GNB24-GNB
02-TGO
21-SEN04-SEN
07-LSO
19-BEN
21-NPL
07-BFA
24-SEN
19-LSO
11-TGO
12-MWI
15-LSO
21-TGO
17-GNB11-GNB
22-TGO
23-MDG
19-NPL
22-RWA11-BFA19-TGO
09-NPL17-SEN04-NER
24-TGO
20-ZMB
15-BFA
12-NPL
22-TZA
24-LSO
04-LSO
19-ZMB
22-LSO19-NER
19-MDG
21-ZMB
07-NPL
04-MLI
04-ZMB
21-RWA
17-NER
04-BFA
17-TGO
22-ZMB
24-NER
04-TGO
23-NPL
21-NER
21-BFA
11-MDG
15-MWI
15-SEN
15-NER08-NPL
22-UGA
02-BEN
19-BFA
19-SEN
02-LSO
10-BFA
09-MLI
17-MDG
21-MLI
11-LSO
12-BFA
15-MDG
10-LSO
19-MLI
15-GNB17-TZA
10-RWA
22-BFA
10-UGA
15-TGO
15-BEN
15-TZA
22-GNB
10-TGO
10-MLI
10-GNB
15-ZMB24-MWI
15-RWA
10-NPL
15-UGA
10-MWI
10-MDG10-NER
10-TZA
10-SEN
10-BEN
050
100
150
200
Tarif
f wat
er
0 .1 .2 .3 .4 .5Trade importance
kernel = epanechnikov, degree = 0, bandwidth = .13
Policy space and trade importance
07-MWI
11-UGA
01-LSO
23-MLI
08-RWA
20-GNB
24-BFA
02-SEN
12-UGA
20-BEN
22-NER
12-SEN
07-NER
02-ZMB
04-TZA24-UGA
20-NPL
08-LSO
08-SEN
04-NPL
07-SEN
18-NPL
17-MLI
01-ZMB
21-TZA
11-SEN
12-RWA
22-MWI19-TZA
20-NER
09-NER
18-ZMB
20-MWI07-MLI11-NPL19-MWI
20-SEN
11-BEN
09-ZMB
22-SEN
21-BEN
22-BEN
11-MWI
17-NPL07-GNB
17-BFA01-NPL
21-MDG
09-LSO
21-GNB
17-MWI
11-TZA
07-ZMB
24-RWA
20-MLI
24-ZMB
02-GNB
20-RWA
23-LSO
04-MDG
17-ZMB
22-NPL
07-RWA
15-MLI
19-RWA
20-BFA
11-ZMB
22-MLI
12-ZMB20-LSO
24-TZA
19-UGA
16-LSO
08-ZMB20-TGO
09-SEN
23-ZMB
21-MWI
10-ZMB
04-GNB
21-UGA
22-MDG
24-NPL
17-LSO
17-BEN
04-MWI
24-MLI
21-LSO
11-MLI
19-GNB
24-GNB
02-TGO
21-SEN
04-SEN
07-LSO
19-BEN
21-NPL
07-BFA24-SEN
19-LSO11-TGO
12-MWI15-LSO
21-TGO
17-GNB
11-GNB
22-TGO
23-MDG
19-NPL
22-RWA
11-BFA
19-TGO
09-NPL
17-SEN
04-NER
24-TGO
20-ZMB15-BFA
12-NPL
22-TZA24-LSO
04-LSO
19-ZMB
22-LSO19-NER
19-MDG
21-ZMB
07-NPL
04-MLI
04-ZMB
21-RWA
17-NER
04-BFA
17-TGO22-ZMB
24-NER
04-TGO
23-NPL
21-NER
21-BFA11-MDG15-MWI
15-SEN
15-NER
08-NPL
22-UGA02-BEN
19-BFA
19-SEN
02-LSO10-BFA
09-MLI17-MDG21-MLI
11-LSO
12-BFA
15-MDG
10-LSO
19-MLI
15-GNB
17-TZA
10-RWA
22-BFA
10-UGA
15-TGO15-BEN
15-TZA
22-GNB
10-TGO10-MLI
10-GNB
15-ZMB
24-MWI
15-RWA
10-NPL
15-UGA
10-MWI10-MDG
10-NER10-TZA
10-SEN
10-BEN
0.2
.4.6
.8m
fn /
boun
d
0 .1 .2 .3 .4 .5Trade importance
kernel = epanechnikov, degree = 0, bandwidth = .08
Policy space used and trade importance
Policy Space in Agricultural Markets 7
050
100
150
200
Tar
iff w
ater
0 .1 .2 .3 .4Trade importance
Tariff water and trade importance
0.2
.4.6
.81
mfn
/ bo
und
0 .1 .2 .3 .4Trade importance
Policy space used and trade importance
Figure 4
Policy Space in Developing non-LDCs, at the HS 2 Digit Level
(a) (b)
(c) (d)
Moreover, also for developing non LDCs there is no clear relationship between policy space and trade
importance. Finally, in the case of non LDCs there are less products concentrated in the top left corner
(for tariff water, which represents low importance and high policy space) than in LDCs. In addition, one
observes that amongst highly imported products with very little policy space in developing countries
there are some that are also very important in terms of value relative to what can be seen for LDCs, as
shown in Figures 3b and 4b. In regard to the use of policy space, there is a substantial use of policy
space in a number of cases (top portions of Figures 4c and 4d), including in products of trade
importance. Still the use of policy space is rather limited in the vast majority of cases. Figure 5 shows
the percentage of imports that are coming from partners of Preferential Trade Agreements (PTA) by
income group, depending on whether they are related to food security or not.
04-COL
16-ARM
20-CUB
24-GTM18-BLZ
12-BWA09-FJI
09-OMN
07-GHA
18-COL
17-BLZ
11-QAT20-VNM
05-PRY07-PRY
08-VEN
18-IDN
09-COL
12-NAM
18-SLV
12-PAN
07-GRD09-BRB
18-GTM
08-KEN
13-SWZ
09-TUR
07-ARM
07-GTM
20-KEN18-BRB
09-KGZ01-ARG18-PAN
07-NIC
08-SLB
17-QAT
04-TUR
17-SLV
07-PER11-FJI
23-SAU
24-COL
02-NIC
15-QAT12-SWZ
02-IDN
11-GHA
04-URY
01-RUS
21-CUB
01-BOL
08-ATG
16-BOL11-URY
15-GRD
07-URY
16-QAT22-CUB
22-SAU
18-BHR11-PER04-ARG
04-KEN
18-SAU23-CIV
20-SLB
16-SWZ11-IDN12-BOL
08-ZAF11-SLV
18-ALB
23-FJI
23-ARE
16-GTM07-ZAF11-PRY
11-VEN
11-VNM
07-SLB02-BLZ
02-BWA
20-VEN19-ARG18-OMN17-VNM17-PAN
15-BHR
16-BRB
19-TUR
19-VEN
07-PAN
04-KGZ
12-ZAF24-SWZ
22-VEN
15-PRY
19-IDN
15-ATG
20-PER
05-VNM
19-CUB09-QAT
17-DMA
08-FJI16-NIC
10-ATG
20-COL
17-NIC19-ZAF
17-ATG
09-SAU
24-ZAF08-PER
18-ARE05-ARG11-SLB
20-ALB16-SLV17-OMN
22-TUR18-DMA
18-ZAF20-ARE
16-DMA
20-ARM17-ARG
12-FJI
04-ALB
12-VEN
11-PAN
17-GTM
07-ALB08-NAM
07-KEN
16-ATG
23-BOL
08-GTM23-KEN
02-PER
07-COL11-KEN
04-PRY18-URY
07-ATG10-DMA
17-PRY23-OMN
09-VNM
16-CUB
16-PAN
09-CUB20-NIC
23-ALB21-VNM
02-BOL
07-TUR20-FJI08-URY10-ARG
08-SWZ08-PAN20-QAT
11-NIC
21-OMN07-SAU
23-BRB21-GHA
20-BOL
01-TUR
12-CUB
20-DMA
18-QAT18-PRY
19-KEN
11-CIV09-ARE24-URY
11-BOL
21-KEN
20-OMN08-BWA
08-BRB
02-COL
19-OMN11-ALB
22-KEN16-GRD
01-ALB
02-ARG
19-ARM22-ARG
19-COL
20-BHR
19-NAM
17-GRD
08-BOL24-BOL07-SLV
09-RUS
13-ARG
09-ARM
20-ZAF
21-ARM
20-ATG
19-VNM
20-BLZ
17-COL
15-ARE
02-NAM
19-GRD
08-SLV
07-NAM
09-ALB08-TUR19-ARE
12-PRY
24-VNM19-FJI
20-GRD
07-SWZ
08-IDN23-GRD
21-ARE
15-VEN07-BRB
20-SAU
15-BRB
08-ARM08-KGZ
15-NAM
20-PRY
19-DMA
07-DMA
08-DMA24-KEN
21-BHR
12-SAU
24-IDN20-SLV
04-VNM
11-GTM
04-BLZ
24-ARG
21-IDN04-BOL
24-TUR
15-VNM
22-PER
19-CIV
07-BWA
18-TUR
22-GTM
08-COL20-GTM
18-RUS
19-GHA17-BRB
19-BHR
20-SWZ
15-SAU
21-ZAF
12-COL
15-BOL07-CUB
22-DMA
21-TUR23-BWA12-ARG
17-ALB
22-MUS
21-SWZ
04-ARM
15-SWZ
21-DMA
24-DMA
21-COL
02-PAN02-SAU
22-COL
22-URY
04-SLB02-SWZ
15-PAN12-RUS
19-BLZ
19-URY21-VEN23-PRY19-BWA
21-ALB
21-FJI07-ARE
15-BWA
19-SAU
10-BOL
04-QAT24-NAM
20-BRB
16-SLB
18-BOL
18-ARM
20-ARG
21-NAM
19-BRB
04-PAN19-QAT
08-ALB21-KGZ
17-VEN
04-GHA
21-SAU17-ARM
19-ATG
22-FJI
11-KGZ
23-SWZ21-GRD
22-GRD
07-VNM
23-VEN
17-BOL
22-VNM15-ALB
02-URY
02-SLV
04-ARE20-NAM08-SAU22-ARM
22-BWA
20-URY
21-BWA
24-ARE
15-CUB
04-GTM
17-URY
23-ARG
21-ATG
15-CIV15-ARM
09-URY
17-FJI02-ALB
22-NIC19-PRY
15-SLB
20-GHA
19-KGZ
20-BWA19-SWZ21-URY
19-SLV23-NAM22-SLV
19-ALB
22-GHA
09-ARG08-ARG17-KGZ04-SAU
24-CIV20-PAN02-FJI
24-BWA
22-BLZ
24-KGZ
15-URY
08-VNM18-KGZ
15-COL
04-NIC23-SLV
21-BRB
23-ZAF19-BOL
12-ARE21-ARG
17-BWA
10-VNM
15-FJI
17-SWZ22-SWZ
15-OMN
02-ZAF
15-KGZ
23-NIC
02-ARE
19-NIC19-GTM
04-SLV
09-KEN
23-CUB
17-NAM02-GHA
24-ALB12-VNM
15-SLV
02-ARM
22-BRB
17-KEN22-PAN10-SLB19-PAN08-ARE
21-GTM
21-BLZ
04-BHR23-GTM
15-GHA
23-PAN
04-VEN
22-ALB
21-PRY21-PAN
21-NIC
04-BRB
01-VEN18-ARG22-BOL23-URY04-CUB17-SLB
21-SLB15-ZAF
15-NIC
02-BRB
24-ARM04-FJI
17-GHA
21-SLV
10-PAN
23-BLZ
15-ARG
02-SLB
10-SLV
15-KEN
02-KGZ10-ALB23-TUR
23-COL
12-TUR02-OMN
02-CUB
02-VNM
15-TUR
10-NIC
19-SLB
10-GTM
22-PRY
10-TUR
04-OMN24-PRY
10-VEN
22-NAM
02-VEN
10-GRD
22-ATG
02-QAT02-RUS
10-GHA
02-DMA
21-BOL
02-ATG
02-GRD
10-CUB
10-COL
10-KEN
10-CIV
050
100
150
200
Tarif
f wat
er
0 .1 .2 .3 .4Trade importance
kernel = epanechnikov, degree = 0, bandwidth = .05
Tariff water and trade importance
12-DMA
04-COL
16-ARM
11-MUS
20-CUB
24-GTM
11-SWZ24-OMN09-MUS
18-BLZ12-BWA09-FJI09-OMN07-GHA18-COL
16-KWT11-DMA
17-BLZ
11-QAT
20-VNM
05-PRY
07-PRY
08-VEN
18-IDN09-COL12-NAM
11-GRD
18-SLV
12-PAN09-BHR07-GRD
09-BRB
18-GTM
05-ZAF
08-KEN13-SWZ
09-TUR
07-ARM
01-PRY
07-GTM20-KEN18-BRB
23-SLB
09-KGZ
01-ARG
18-PAN
07-NIC
08-SLB17-QAT
04-TUR
11-ATG
15-IDN
17-SLV
07-PER
11-BRB
11-FJI
23-SAU
24-COL
02-NIC15-QAT12-SWZ
02-IDN
01-MUS
11-GHA
04-URY
01-RUS
21-CUB
01-BOL
08-ATG16-BOL
11-URY15-GRD
23-KWT
07-URY
16-QAT
22-CUB
22-SAU
18-BHR11-PER
09-KWT
04-ARG
12-KWT
04-KEN
01-SAU
18-SAU
23-CIV
20-SLB
16-SWZ
11-IDN
12-BOL
08-ZAF11-SLV
18-ALB
23-FJI
23-ARE
16-GTM07-ZAF11-PRY
11-VEN
05-MUS
11-VNM
07-SLB
02-BLZ
01-ZAF
02-BWA
20-VEN19-ARG
18-OMN
17-VNM
17-PAN
15-BHR
16-BRB
19-TUR
19-VEN07-PAN
04-ZAF
04-KGZ
12-ZAF
01-IDN
22-OMN
24-SWZ
22-VEN
15-PRY
19-IDN
15-ATG20-PER
18-MUS
05-VNM
19-CUB
11-ZAF09-QAT
17-DMA
08-FJI
16-NIC
11-BLZ
10-ATG20-COL
12-KEN
17-NIC
19-ZAF
17-ATG09-SAU
24-ZAF
08-PER
18-ARE
16-MUS
05-ARG
11-SLB
20-ALB
16-SLV
17-OMN22-TUR
18-DMA
18-ZAF
09-BLZ
20-ARE
16-DMA12-NIC
20-ARM
12-GTM09-NAM
17-ARG
12-FJI
04-ALB
12-VEN11-PAN
17-GTM
07-ALB
08-NAM
07-KEN
16-ATG
23-BOL
23-GHA
08-GTM23-KEN18-KWT
02-PER07-COL
11-KEN
04-PRY
18-URY
07-ATG10-DMA
17-PRY
23-OMN
09-VNM
16-CUB
16-PAN
09-CUB
20-NIC
23-ALB
21-VNM
02-BOL
07-TUR20-FJI
08-URY
10-ARG
11-NAM11-BWA09-ZAF
08-SWZ
08-PAN
20-QAT
12-URY
11-NIC
21-OMN
07-SAU23-BRB21-GHA
20-BOL01-TUR
12-CUB
20-DMA
18-QAT
17-KWT09-SWZ
18-PRY
19-PER
19-KEN
11-CIV
09-ARE
24-URY
11-BOL
21-KEN24-NIC
20-OMN
08-BWA
08-BRB
02-COL
19-OMN
11-ALB
22-KEN
08-MUS
16-GRD
01-ALB
02-ARG
19-ARM22-ARG
19-COL
01-KWT
20-BHR
19-NAM
17-GRD08-BOL
17-PER
24-BOL
07-SLV
23-MUS
09-RUS
13-ARG
09-ARM
09-BWA
20-ZAF
21-ARM
20-ATG
19-VNM
20-BLZ
17-COL
15-ARE
02-NAM19-GRD
08-SLV
07-NAM
20-MUS
09-ALB08-TUR
19-ARE
12-PRY
22-SLB
07-MUS
07-IDN
08-OMN
24-VNM
19-FJI
22-BHR24-QAT20-KWT
20-GRD
07-SWZ
15-KWT
08-IDN
07-OMN12-PER
23-GRD
21-ARE
15-VEN
07-BRB20-SAU
15-BRB
08-ARM
08-KGZ
15-NAM
24-KWT
20-PRY
17-BHR
19-DMA
07-DMA
08-DMA
24-KEN
21-BHR12-SAU
24-IDN
20-SLV04-VNM
11-GTM04-BLZ
17-MUS
24-ARG21-IDN
04-BOL
24-TUR
15-VNM
22-PER
19-CIV
07-BWA
18-TUR21-KWT
22-GTM
08-COL20-GTM
23-ARM
18-RUS
19-GHA
21-QAT
17-BRB
19-BHR
20-SWZ15-SAU
21-ZAF12-COL
15-BOL
07-CUB
22-DMA
21-TUR
23-BWA
04-PER
12-ARG
17-ALB
22-QAT
22-MUS
21-SWZ
04-ARM
15-SWZ21-DMA
04-NAM
24-DMA
21-COL
16-BLZ
02-PAN
02-SAU
22-COL
22-URY
04-SLB
02-SWZ15-PAN
02-GTM
12-RUS
19-BLZ
19-URY21-VEN
23-PRY
19-BWA
21-ALB
21-FJI
08-QAT
07-ARE
08-BHR
15-BWA
19-SAU
21-PER15-BLZ
10-BOL04-QAT
24-NAM
20-BRB16-SLB
18-BOL
18-ARM
20-ARG
17-SAU
21-NAM
19-BRB
04-PAN
21-MUS
19-QAT
08-ALB
21-KGZ
17-VEN04-GHA
02-MUS07-BHR
21-SAU
19-MUS19-KWT
17-ARM
19-ATG
22-FJI
11-KGZ23-SWZ
21-GRD22-GRD
07-VNM
23-ATG
23-VEN17-BOL
22-VNM15-ALB
02-URY
02-SLV
04-ARE
20-NAM
15-MUS
08-SAU22-ARM22-BWA
07-KWT12-BRB
20-URY
08-KWT
21-BWA
24-ARE15-CUB04-GTM
17-URY
23-ARG21-ATG
15-CIV
15-ARM
09-URY
17-FJI
02-ALB
04-KWT
22-NIC
19-PRY
04-ATG04-BWA
15-SLB
20-GHA
19-KGZ
20-BWA
01-QAT
19-SWZ21-URY
19-SLV
23-NAM
22-SLV
19-ALB
22-ARE
22-GHA
04-IDN24-MUS
09-ARG
17-ZAF
08-ARG
17-KGZ
04-SAU
24-CIV
20-PAN02-FJI
24-BWA
22-BLZ
24-KGZ
15-URY
08-VNM
18-KGZ
15-COL
04-NIC
07-QAT
23-SLV
04-SWZ10-BHR12-IDN
21-BRB23-ZAF
19-BOL
17-ARE15-GTM
12-ARE
04-DMA10-ARE
21-ARG
17-BWA
10-VNM
07-FJI
15-FJI
17-SWZ22-SWZ
15-OMN
02-ZAF
15-KGZ
23-NIC02-ARE
19-NIC
04-GRD
19-GTM
04-SLV
09-KEN23-CUB
22-ZAF
17-NAM
02-GHA
24-ALB
12-VNM
10-BLZ
15-SLV
02-ARM
22-BRB
10-BRB
17-KEN
10-PRY
22-PAN
10-SLB
19-PAN
08-ARE
23-VNM
21-GTM
10-NAM
21-BLZ04-BHR23-GTM15-GHA23-PAN
04-VEN
22-ALB
21-PRY
21-PAN21-NIC04-BRB
15-PER
01-VEN18-ARG
22-BOL
17-IDN
23-URY
04-CUB17-SLB
10-KGZ
21-SLB15-ZAF15-NIC
02-BRB24-ARM
04-FJI
17-GHA
21-SLV
10-PAN
23-BLZ
04-MUS
15-ARG
10-QAT10-URY
02-SLB
10-SLV
15-KEN
02-KGZ
10-ALB
23-TUR
23-PER
23-COL
12-TUR
02-OMN
02-CUB
02-VNM
15-TUR
10-NIC
10-ARM10-BWA10-ZAF10-SWZ
19-SLB10-GTM24-BHR02-KWT
22-PRY
10-TUR
23-IDN10-MUS
04-OMN
24-PRY
10-VEN
22-NAM
10-KWT15-DMA02-BHR
02-VEN
10-GRD
24-BLZ
10-IDN
10-FJI
22-ATG02-QAT
10-OMN
02-RUS
10-GHA02-DMA
21-BOL
02-ATG02-GRD
10-CUB
10-PER
10-COL10-KEN
10-SAU
10-CIV
0.2
.4.6
.81
mfn
/ bo
und
0 .1 .2 .3 .4Trade importance
kernel = epanechnikov, degree = 0, bandwidth = .06
Policy space used and trade importance
8 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
Figure 5
Share of imports from partners within RTAs, by income level
What is considered important from a food security perspective in the present study is cereals and oils,
fats as well as sugars, which correspond to HS 2 digit categories 10, 15 and 17. The main difference
can be seen for LDCs for which there is considerably more imports from partners of PTAs in products
that are not related to food security. This may shed some light on one of the differences found in the
empirical section between the coefficients on food security for tariff water and true tariff water, which
takes into account preferential trade.
Figure 6 looks at the MFN/bound and tariff water from 2008 to 2013 for all three income level groups.
One can see that there hasn't been a drastic change of the space between the MFN applied and the
bound rate over the last five years except for the large decrease of tariff water between 2008 and 2009
in LDCs, most likely reflecting a wave of protectionism due to the financial and economic crisis.
0.1
.2.3
.4
Developed Developing LDCs
Share of imports from partners of RTAs
No food security Food security
Policy Space in Agricultural Markets 9
Figure 6
Policy space
Figure 7 then shows the same two variables separated into products that were or that were not
produced in the country during that year.
Figure 7
Policy space for imported and produced goods
It is very important to bear in the mind the descriptive aspect of this figure as opposed to a causal one.
As will be discussed in the empirical section, production is an endogenous variable. Local production
may be a reason for governments to set higher MFN applied rates and reversely higher MFN rates will
reduce foreign competition and enable producers to have larger shares of the market. Nonetheless, by
separating into two categories of produced versus non-produced locally, at a relatively aggregated
level, one can see the general level of MFN/bound and tariff water in both groups. This isn't looking at
the level of production which of course would be influenced by the MFN applied rate. Reverse causality
may be taking place if sufficiently low MFN applied rates are enabling high enough imports to
completely exclude local production across a whole HS 2 digit line. Even though this is possible, it may
possibly be assumed that it is not frequent. What shows up in the figure is that except for LDCs in 2008
and for developing countries in 2012 the MFN/bound was sytematically lower when the country
produced the product. This however did not tend to be the case in developed countries. The outcome
for tariff water is similar for developing countries, and the same pattern now sticks out for developed
countries too. However, this no longer seems to be the case for LDCs. One may however argue that
the ratio of the MFN/bound better reflects the use of policy space, as the value is not absolute but
relative.
In general there doesn't tend to be a clear relationship between the importance of the products in terms
of imported value and policy space. Policy space (related to tariffs) is clearly very limited only in relation
020
4060
8010
0
2008 2009 2010 2011 2012 2013
Tariff water at the HS 2 digit level(weighted average over HS 2 digit lines)
LDCs DevelopingDeveloped
0.2
.4.6
.8
2008 2009 2010 2011 2012 2013
mfn / bound at the HS 2 digit level(weighted average over HS 2 digit lines)
LDCs DevelopingDeveloped
050
100
150
2008 2009 2010 2011 2012
Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp
Tariff water at the HS 2 digit level(weighted average over HS 2 digit lines)
LDCs DevelopingDeveloped
0.2
.4.6
.8
2008 2009 2010 2011 2012
Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp Imp + prod Only imp
mfn / bound at the HS 2 digit level(weighted average over HS 2 digit lines)
LDCs DevelopingDeveloped
10 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
to developed countries' imports. For LDCs there is virtually no imports for which policy space is not
available. In general, policy space for developing countries is often available but to a different degree
depending on specific products and countries. An important question that emerges from this analysis is
why policy space seems to be used in some cases but not in others. This may be due to the protection
of domestic production, food security and participation in global value chains. This question is
addressed in the empirical section.
4. EMPIRICAL SPECIFICATION
The first step is to run regressions on tariff water and available space at the HS 2 digit level using the
two measures presented in section 3. They are complemented by two measures that take into account
the presence of prohibitive tariffs that are below the bound rate and for the presence of preferential
tariffs. This is consistent with studies such as Foletti et al. (2011), and the reason is that countries won't
increase their MFN applied rate above the prohibitive rate. This means that the implicit bound is the
prohibitive rate and not the official bound rate in these cases. The preferential rate is below the MFN
applied rate and the country is tied to its contractual agreements, making increases of these rates
particularly complicated.
These two extra measures used are called true available space and true tariff water and are presented
in equations (3) and (4). The true tariff water uses the idea of dammed water found in Foletti et al. (2011),
where the prohibitive tariff is used when it is below the bound rate instead of using the bound rate itself,
and where the whole expression is multiplied by the share of imports taking place under a PTA. This
will therefore reduce tariff water if there is at least some trade taking place under a PTA, and reduces it
to zero in the most extreme case where all trade takes place under a PTA. The same idea is used to
create true available policy space. In this case the prohibitive rate is also used in place of the bound
rate when relevant, and the whole expression is multiplied by 1 minus the share of imports taking place
under a PTA in order to give us the loss of space due to PTAs. This loss of space is added to the initial
available space (including the prohibitive rates). This will lead to a higher value, meaning less available
space, with the extreme case being that the true MFN/bound is equal to 1 if all trade takes place under
a PTA.
��4�� ����5 �����(,�,6 =∑ ���",�,7∗ ��",8,�,7∑ ���",�,7∗ ��",8,�,78∋9 :min:$��(,>,�,6; @��(,>,�,6A − ���(,>,�A ∗(1 − BCDE��",8,�,7 ��",8,�,7>∋F )(3)
��4� * �+ ,+�-. /���(,�,6 = H 2��(,�,6 ∗ 3��(,>,�,6∑ 2��(,�,6 ∗ 3��(,>,�,6>∋F I ���(,>,�,6min:$��(,>,�,6; @��(,>,�,6AJ>∋F
+
H 3@�����(,>,�,63��(,>,�,6 ) ∗ I1 −H 2��(,�,6 ∗ 3��(,>,�,6∑ 2��(,�,6 ∗ 3��(,>,�,6>∋F I ���(,>,�,6min:$��(,>,�,6; @��(,>,�,6AJ>∋F J>∋F (4)
2��(,� is the import demand elasticity of a given HS 6 digit product in country c, 3��(,>,� are the imports of country c from country d for the same product, $��(,>,� is the bound rate, ���(,>,� the MFN applied tariff , and finally @��(,>,� the level of the prohibitive tariff, computed following Foletti et al. (2011), with @��(,>,� = ���(,>,� +MN0��",8,����",� .
Policy Space in Agricultural Markets 11
The mean levels of elasticity within a given HS 2 digit line are taken by weighting by the value of
imports and the share of intermediates is computed in the same way. For the two explained variables,
namely tariff water and available space, the mean is computed as in equations (5) and (6), using
imports as well as the elasticity of import demand as weights, this time aggregating over HS 6 digit line
products also.
��4�� ����5 ������,�,6 =∑ 6COD6PC%EEQP6DC��",�,7∗���",�,7∗ ��",�,7∑ ���",�,7∗ ��",�,7��"∋�����(∋��� (5)
��4� * �+ ,+�-. /����,�,6 =∑ 6CODPRP%SPTSD�UP�D��",�,7∗���",�,7∗ ��",�,7∑ ���",�,7∗ ��",�,7��"∋�����(∋��� (6)
True tariff water will be bound between 0 and the classical tariff water measure. Concerning true
available space, its value will be between the classical available space and 1. These two variables will
be the explained variables, as shown in equation (7) which shows the empirical specification used:
V���,�,6 = W + X2���,�,6 +Y�Z���[�\� ��-��� + ]�^^\-�/4���_��� + `.�^\4/��^Z��� + a�,6 +b���,�,6 (7)
where V���,�,6 will either be tariff water or available space. 2���,� is as before the import demand elasticity of a given HS 6 digit product in country c which is aggregated to the HS 2 digit level using import
values as weights. Intermediates is the trade weighted share of HS 6 digit products that are used as
intermediates in the production process within a given HS 2 digit product category. Food security is a
dummy variable that takes a value of 1 if the product, as mentioned above, is considered important
from a food security perspective, namely cereals and oils, fats as well as sugars, which correspond to
HS 2 digit categories 10, 15 and 17. Production is a dummy variable that takes a value of 1 if the good
was produced in the country in the same year and 0 otherwise. The choice of a dummy rather than the
value of production is clearly preferred in order to diminish the problematic endogeneity of the latter,
without obviously overcoming it entirely. It is unlikely that there isn't at least one producer who stays on
the market in the case of an MFN applied decrease. New producers starting to produce a product
following an increase of the MFN applied will certainly happen, but it would be quite rare at that level of
aggregation to observe a change on a whole HS 2 digit line. Finally, a�6 are country-year fixed effects and b���,�,6 is the error term. As the fixed effects are interacted, they not only control for anything specific to a given year or country, but also anything that is specific to a country in a given year. This
controls for the macroeconomic variables mentioned above, amongst which one can mention GDP,
inflation, the exchange rate and fiscal needs. Their expected influence on tariffs is reviewed below so
as to have a comprehensive view of what is influencing tariffs. Increased GDP will tend to increase
overall demand, therefore impacting consumer surplus. Inflation will tend to decrease producers' profits
due to products being relatively more expensive with respect to imports. This will in turn influence
demand for imports, increasing tariff revenue and influencing consumer surplus. The exchange rate will
change the price of imports, therefore influencing consumer surplus, tariff revenue due to a change in
the imported value and/or volume, and also producer profits due to a change in competitiveness. A
country's fiscal needs will influence the level of tariff revenue to aim for.
The interest for the food security variable and intermediates doesn't enable us to use item specific fixed
effects. Results of Ordinary Least Squares Regressions (OLS) run on equation (7) are presented in
Table 1. One must bear in mind that the OLS estimations for available space are biased due to the fact
that the explained variable is a fraction. This issue is explained and treated below, but the results of the
OLS regression are still presented as a reference in Table 1. Papke and Wooldridge (1996, 2008) show
why the bounded nature of a fraction and the possibility of observing values at the boundaries causes
estimation issues. More specifically, the effect of a given explanatory variable cannot be constant over
the whole range of values that it can take. Bluhm (2013) proposes a QMLE Stata routine entitled
12 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
fhetprob to perform a fractional probit estimation with heteroskedasticity based on the work of Papke
and Wooldridge (2008).7 In the present case, as in Bluhm (2013), time averages of all explanatory
variables are included in order to account for unobserved heterogeneity in the form of Correlated
Random Effects (CRE). The length of each spell is also included in the estimation and this is considered
potentially endogenous. Spell lengths of 1 must be dropped in order to avoid perfect collinearity
between the variable and its mean, as explained in Wooldridge (2008). Year dummies are also included
and errors are clustered by product. The results of this procedure are presented in columns 5 and 6 of
Table 1.
Table 1
OLS and fractional heteroskedastic probit
water
true
water MFN/bound
true
MFN/bound MFN/bound
true
MFN/bound
OLS OLS Fractional Heterosk. Probit
elasticity 0.006 0.055*** -0.000 -0.003*** -0.001 -0.024*
(0.24) (5.76) (-1.10) (-7.71) (-0.60) (-1.70)
intermediate 2.333*** -3.191*** -0.169*** -0.209*** -0.237*** -0.284***
(4.97) (-20.47) (-32.50) (-30.89) (-3.09) (-3.22)
food security 3.173*** 0.786*** 0.014** 0.014* 0.068 0.027
(4.33) (3.87) (2.24) (1.83) (0.50) (0.22)
production dummy
-0.890* -0.366** -0.005 -0.013** -0.002 0.137***
(-1.76) (-2.11) (-1.00) (-1.99) (-0.13) (4.58)
constant 1.206* 1.668*** 0.807*** 1.024*** -1.890 -1.750***
(1.87) (5.92) (17.94) (53.46) (-1.57) (-4.61)
country x year dummies
Yes Yes Yes Yes No No
year dummies Yes Yes Yes Yes Yes Yes
spell dummies No No No No Yes Yes
N 9774 9676 9774 9676 9601 9502
R2 0.783 0.437 0.594 0.451
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
The results discussed below look both at tariff water and use of policy space, for which the reference is
the fractional heteroskedastic probit. Results are mostly consistent with the OLS case but there are
some exceptions and as discussed higher up the OLS is potentially biased and is shown for
transparency and reference. As can be seen in Table 1, the elasticity does not seem to play a role in the
setting of tariff water or the MFN/bound (for which the term use of policy space is used below).
However, the effect for true tariff water is positive and significant and for the use of policy space it is
negative and significant. One must remember that higher values of tariff water are associated with
7 Papke and Wooldridge (1996) put forward that this issue is very similar to the case of models with binary data. They propose, based on Gourieroux, Montfort, and Trognon (1984) as well as McCullagh and Nelder (1989) to use Bernouilli Quasi Maximum Likelihood Estimator (QMLE), arguing that it is easy to maximize and is a member of the linear exponential family and consistent. Papke and Wooldridge (2008) note that when in the presence of panel data, there is an extra problem as one must be sure that standard errors are robust to arbitrary serial correlation and on top of that one must address the fact the fact that the explanatory variables may be correlated to unobserved heterogeneity. They suggest the use of a probit rather than an logit estimator, which even though often very similar, has the advantage of better handling endogenous variables. The method they propose is however not adapted to unbalanced panel data. This issue is solved in
Bluhm (2013), who proposes a QMLE Stata routine entitled fhetprob to perform a fractional probit estimation with heteroskedasticity based on the work of Papke and Wooldridge (2008). Bluhm (2013) explains that the conditional variance has to be able to vary with the nature of the unbalancedness, therefore requiring heteroskedasticity in the model.
Policy Space in Agricultural Markets 13
lower ratios of the use of policy space, such that opposite signs on the coefficients of both variables
are consistent. The coefficients on intermediates are positive for tariff water and negative for the use of
policy space. The latter is also negative when considering the alternative true measure of the use of
policy space. It is negative for true tariff water, which contrasts with the other coefficients. Reasons for
this could be lower prohibitive rates in intermediate goods or more trade in intermediates with
preferential trade agreement partners. Except for this, results for intermediates are consistent with the
prior that the policymaker is keeping the price of inputs low to favor local production of processed
products. Products that are important in terms of food security tend to have more tariff water and true
tariff water. The use of policy space is not significantly affected by this variable in the fractional
heteroskedastic probit regressions, which is the preferred method due to the potential bias using OLS.
The coefficients on the production dummy tell us that when the country is producing the good there is
less tariff water and less true tariff water, for the use of policy space there is no significant effect and for
the true use of policy space it is positive (in the fractional heteroskedastic probit specification). This
suggests some protection of local producers.
In order to determine whether the effects of the different variables vary by income levels, regressions are run on three different samples, one for each income level. Even if there aren't necessarily strong
priors on the differences one may expect, reasons for believing that there may be differences include the fact that wealthier governments may have more technical skills to set MFN applied levels according to given economic conditions and item specificities, and poorer countries may give more importance to
issues such as food security. Results of these regressions are given in tables 2 to 4.
Table 2
Water and true water by income group (OLS)
Water
LDCs
Water
Developing
Water
Developed
True water
LDCs
True water
Developing
True water
Developed
elasticity 0.096*** -0.008 0.032 0.207*** 0.053*** 0.010***
(2.78) (-0.29) (1.20) (4.93) (5.06) (3.40)
intermediate 7.623*** 0.720 1.586 -8.026*** -2.045*** -0.162
(6.22) (1.33) (1.39) (-19.49) (-12.11) (-0.99)
food security -4.614*** 5.216*** 7.346*** 0.813* 0.932*** -0.257
(-2.60) (6.06) (3.67) (1.68) (3.91) (-0.91)
production dummy -4.540*** -0.145 2.211 -1.255** -0.118 -0.157
(-3.47) (-0.25) (1.62) (-2.22) (-0.69) (-1.03)
constant 51.930*** 1.286* 5.126*** 33.770*** 1.154*** 1.214***
(6.78) (1.76) (4.66) (6.03) (5.38) (8.05)
country x year dummies
Yes Yes Yes Yes Yes Yes
N 1965 6781 756 1965 6781 756
R2 0.734 0.744 0.445 0.316 0.434 0.129
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
14 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
Table 3
MFN/bound and true MFN/bound by income group (OLS)
MFN/bound
LDCs
MFN/bound
Developing
MFN/bound
Developed
True
MFN/bound
LDCs
True
MFN/bound
Developing
True
MFN/bound
Developed
elasticity -0.001** -0.000 0.001 -0.007*** -0.003*** 0.001 (-2.43) (-1.04) (0.90) (-4.31) (-6.84) (0.77)
intermediate -0.163*** -0.176*** -0.158*** -0.112*** -0.239*** -0.235*** (-21.44) (-26.13) (-7.10) (-9.06) (-28.56) (-8.49)
food security 0.054*** 0.004 0.008 0.030** 0.008 0.044 (4.22) (0.49) (0.40) (2.12) (0.81) (1.51)
production dummy
-0.010 -0.005 -0.017 -0.018 -0.011 -0.020 (-1.11) (-0.75) (-0.77) (-1.49) (-1.37) (-0.77)
constant 0.375*** 0.820*** 0.307*** 0.419*** 1.035*** 0.687*** (4.99) (17.98) (8.88) (2.58) (49.88) (10.44)
country x year dummies
Yes Yes Yes Yes Yes Yes
N 1965 6781 756 1965 6781 756 R2 0.415 0.578 0.675 0.205 0.483 0.510
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
Table 4
MFN/bound and true MFN/bound by income group (Fract. Heter. Probit)
MFN/bound
LDCs
MFN/bound
Developing
MFN/bound
Developed
True
MFN/bound
LDCs
True
MFN/bound
Developing
True
MFN/bound
Developed
elasticity -0.012*** 0.001 -0.017*** -0.055 -0.022 -0.010*** (-3.47) (1.30) (-3.25) (-1.56) (-1.44) (-2.77)
intermediate -0.262* -0.242*** -0.647 -0.176 -0.477*** -1.016 (-1.72) (-2.64) . (-1.30) (-3.62) (-1.62)
food security 0.202 0.024 -0.141 0.120 -0.027 -0.027 (1.60) (0.19) (-0.68) (0.98) (-0.21) (-0.18) production dummy
-0.017 -0.007 0.160** 0.081* 0.289*** 0.307*
(-0.94) (-0.30) (2.09) (1.73) (6.67) (1.78)
constant -1.801*** -1.475*** -6.009*** -0.404 -1.144*** -5.430***
(-3.11) (-3.63) (-4.67) (-1.07) (-3.59) (-3.76)
year dummies Yes Yes Yes Yes Yes Yes spell dummies Yes Yes Yes Yes Yes Yes
N 1982 6859 760 1965 6781 756
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
The elasticity of import demand coefficient for tariff water is positive and significant only for LDCs, but
for true water it is positive and significant for all income groups. The effect remains the largest in LDCs,
followed by developing and developed countries. When looking at the use of policy space (once again
in the fractional heteroskedastic probit specification) the effect is negative for LDCs and developed
countries and for the true use only in developed countries. As was shown in the descriptive statistics,
there tends to be considerably more water in LDCs, in particular due to high bound rates. This may
explain this pattern in which the effect on water sticks out for LDCs whereas it is more the case for the
use of policy space in developed countries. The coefficient on intermediates is only positive in LDCs,
and as in the regression with all regions together, this sign switches for true water and it is also
negative for true water in developing countries. It is negative for the use of policy space in both these
Policy Space in Agricultural Markets 15
income regions and only significant and negative for its true counterpart in developing countries. The
coefficient on food security indicates that there is less tariff water in products related to food security in
LDCs but more water in developed and developing countries. The sign changes for LDCs when
considering true tariff water, which can have two potential explanations, on one hand prohibitive rates
may tend to be higher in food products due to the goods being of primary necessity, and on the other
hand, as illustrated in Figure 5 and discussed in section 3, there is more trade taking place with PTA
partners in non-food security related products in LDCs. For other income groups the difference is only
very slight, and in the opposite way. It could help explain the coefficient for developed countries going
from positive and significant to non-significant. Food security doesn't seem to be related to the use of
policy space, probably meaning that for food security products it is the absolute levels of the MFN
applied and the bound rates that are playing a role. Finally, the production dummy coefficient tells us
that tariff water and true tariff water is lower in LDCs when the good is also produced by the country
but that there tends to be no difference for the two other income groups. For the true use of policy
space it is positive in all income groups.
5. ROBUSTNESS CHECKS
Regressions are run at a lower level of aggregation to check the consistency of results at the cost of
losing the production dummy. Results for regressions run at the HS 6 digit line level are shown in Table
5. Nearly all coefficients are consistent. There are two cases where the confidence level changes but
the coefficient remains significant and still of the same sign, namely food security in the OLS regression
on the true MFN/bound (column 4) and the import demand elasticity in the fractional heteroskedastic
regression once again for the true MFN/bound (column 6). The same coefficient but for the classical
MFN/bound goes from not significant in the HS 2 digit case to negative and significant (consistent with
the true MFN/bound coefficient) in the HS 6 digit case. Finally, food significance went from being non-
significant for the HS 2 digit to positive and significant for the HS 6 digit true MFN/bound specification
(column 6).
Table 5
OLS and fractional heteroskedastic probit at the HS 6 digit level
water true water MFN/bound
true
MFN/bound MFN/bound
true
MFN/bound
OLS OLS Fractional Heterosk. Probit
elasticity -0.004 0.034*** 0.000 -0.001*** -0.005*** -0.009** (-1.03) (22.86) (0.55) (-15.40) (-4.25) (-2.17) intermediate 3.303*** -2.606*** -0.120*** -0.138*** -0.315*** -0.445*** (15.20) (-32.66) (-64.26) (-55.54) (-8.17) (-9.69) food security 3.097*** 0.766*** 0.005** 0.029*** -0.002 0.117** (13.07) (6.18) (2.16) (8.95) (-0.06) (2.1) constant 0.511** 1.800*** 0.848*** 0.949*** -0.335*** -2.151*** (2.16) (10.26) (54.97) (82.60) (-6.45) (-3.86)
country x year dummies
Yes Yes Yes Yes No No
year dummies Yes Yes Yes Yes Yes Yes spell dummies No No No No Yes Yes
N 66801 62898 66801 62898 65129 60952 R2 0.623 0.306 0.558 0.407
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
16 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
Table 6
OLS and fractional heteroskedastic probit excluding the import demand elasticity
water true water MFN/bound
true
MFN/bound MFN/bound
true
MFN/bound
OLS OLS Fractional Heterosk. Probit
intermediate 2.330*** -3.221*** -0.169*** -0.208*** -0.239*** -0.281***
(4.96) (-20.58) (-32.49) (-30.57) (-3.14) (-2.95)
food security 3.175*** 0.806*** 0.014** 0.013* 0.069 0.024
(4.32) (3.96) (2.23) (1.71) (0.52) (0.21)
production dummy -0.892* -0.394** -0.005 -0.011* -0.004 0.131***
(-1.77) (-2.28) (-0.98) (-1.77) (-0.23) (4.51)
constant 1.201* 1.616*** 0.807*** 1.027*** -1.987** -1.852***
(1.86) (5.76) (17.95) (53.36) (-2.06) (-5.22)
country x year dummies
Yes Yes Yes Yes No No
year dummies Yes Yes Yes Yes Yes Yes
spell dummies No No No No Yes Yes
N 9774 9676 9774 9676 9601 9502
R2 0.783 0.433 0.594 0.446
t statistics in parentheses * p<0.10, ** p<0.05, *** p<0.01
Another concern that one may have is the use of the import demand elasticity as an explanatory
variable. The reason for this is that it is also used in the weighting of the explained variables. Table 6
shows the main results run without this variable. All results are consistent with coefficients that are all
of the same sign and level of significance in all specifications, except for the production dummy in the
OLS regression run on the true use of policy space (column 4), for which the confidence level goes
from 5 percent to 10 percent.
Policy Space in Agricultural Markets 17
6. CONCLUSION
This paper investigates the extent of policy flexibility and its use in relation to tariff setting in agricultural
products. To this end, the magnitude of flexibility - the policy space countries have under international
commitments - is measured by the difference between the bound rate and the MFN applied rate. The
use of policy space is measured by the ratio of the MFN applied to the bound rate. Using econometric
methods this study analyzes whether policy space is influenced by four specific factors: the elasticity of
import demand of the product, the use of the product as an intermediate good, whether the product is
important for food security, and whether the product is also domestically produced.
A general finding is that policy space in agricultural products is generally available, and only limited for
developed countries. Many developing countries have ample room to raise tariffs in most agricultural
imports without infringing binding commitments. For LDCs there is virtually no imports for which policy
space is not available. The findings indicate that four specific factors are related to the use of policy
space. In particular, policy space tends to be used relatively less for products with lower elasticity of
demand. This is consistent with relatively higher rates of protection on elastic products. The results also
find that policy space is seldom used for intermediate products. This may suggest that processing
industries are lobbying governments to keep taxation relatively lower on intermediate products. In
regard of products relevant for food security, the results find that policy space is larger but that there is
no difference in its use. This suggests two things. First, governments may be aiming for access to
cheaper food products, therefore helping consumer welfare. Second, governments may retain policy
space so as to increase the MFN applied in case of need. In regard to products that face domestic
competitors, the results indicate lower tariff water and more use of policy space, suggesting that
producer protection is an issue related to the level of policy space to use and the level of market
protection to set. When looking at the results for different country groupings, it appears that for LDCs,
the overall results are similar to the results with all income groups. They even tend to be larger in
magnitude for the availability of policy space. Results for developing countries indicate that although
the four main factors still play a role, they do so to a lesser extent. For developed countries there is a
similar tendency, despite intermediates no longer playing a significant role.
Overall, the main message of this paper is that most developing countries retain a large degree of
policy space as the MFN applied rates are usually well below the bound rates. Policymakers seem to
be basing the choice of the applied tariffs on a number of product-specific variables that seemingly
correspond to a complex mixture of optimization of consumer surplus, producer profits and fiscal
needs, all this associated with the fact that governments preferably want to have some available space
in order to be able to adjust to any future economic shocks. The impact of the different variables on
policy space seems stronger in LDCs, but especially when considering tariff water. This tendency is
less pronounced for the use of policy space. As a final caveat, it is important to underline that the
analysis does not take into account any policy restrictions on the use of non-tariff measures. Indeed,
one interesting path for future research would be to explore whether policy space is limited by some of
these types of trade policy instruments.
18 POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES
BIBLIOGRAPHY
Amador, M. and K. Bagwell (2012), "Tariff Revenue and Tariff Caps", American Economic Review,
102(3), 459-65.
Bacchetta, M. and Piermartini, R. (2011), "The Value of Bindings", WTO staff working paper.
Bluhm, R. (2013), "fhetprob: A fast qmle stata routine for fractional probit models with multiplicative Heteroskedasticity", Forthcoming as UNU-MERIT working paper.
Estevadeordal, A., C. Freund and E. Ornelas (2008), "Does regionalism affect trade liberalization toward
nonmembers?", The Quarterly Journal of Economics, 123(4), 1531-75.
Foletti, L., M. Fugazza, A. Nicita and M. Olarreaga (2011), "Smoke in the (Tariff) Water", The World
Economy, 34(2), 248-64.
Francois J. and W. Martin (2004), "Commercial Policy Variability, Bindings and Market Access",
European Economic Review, 48(3), 665-79.
Gourieroux, C., A. Monfort and A. Trognon (1984), "Pseudo-maximum likelihood methods: theory",
Econometrica, 52(3), 681-00.
Handley, K. (2014), "Exporting under trade policy uncertainty: Theory and evidence", Journal of
International Economics, 94(1), 50-66.
Horn, H., G. Maggi and R. Staiger (2010), "Trade Agreements as Endogenously Incomplete Contracts",
American Economic Review, 100(1), 394-419.
Humphrey, J. and Memedovic, O. (2006), "Global Value Chains in the Agrifood Sector", UNIDO Working Paper Series, UNIDO Strategic Research and Economics Branch, United Nations
Industrial Development Organization.
Kee, H.L., A. Nicita and M. Olarreaga (2009), "Estimating Trade Restrictiveness Indices", Economic
Journal, 119(534), 172-99.
McCullagh, P. and J. A. Nelder (1989), Generalized Linear Models, 2nd edition, Chapman and Hall, New
York.
Multilateral Trade Negotiations on Agriculture - A Resource Manual (2000), Part 2 - Agreement on
Agriculture, Module 2 - Preparing for Negotiating Further Reductions of the Bound Tariffs, R. Sharma - Commodities and Trade Division, Food and Agriculture Organization of the United Nations, Rome.
Organisation for Economic Co-operation and Development (2012), "Mapping Global Value Chains",
Policy Dialogue on Aid for Trade, Trade and Agricultural Directorate and Trade Committee, The OECD Conference Centre, Paris.
Papke, E. and J. Wooldridge (1996), "Econometric Methods for Fractional Response Variables with an
Application to 401 (K) Plan Participation Rates", Journal of Applied Econometrics, 11(6), 619-32.
Papke, E. and J. Wooldridge (2008), "Panel Data Methods for Fractional Response Variables with an
Application to Test Pass Rates", Journal of Econometrics, 145(1-2), 121-33.
Sala D., Schroder P.H. and E. Yalcin (2009), "Market Access through Bound Tariffs", Scottish Journal of
Political Economy, 57(3), 272-89.
U n i t e d n at i o n s C o n f e r e n C e o n t r a d e a n d d e v e l o p m e n t
POLICY SPACE IN AGRICULTURAL MARKETS
POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIESRESEARCH STUDY SERIES No. 73
Printed at United Nations, Geneva1600845 (E) – January 2016 – 245
UNCTAD/ITCD/TAB/75
United Nations publicationISSN 1607-8291