WPS5200
Policy Research Working Paper 5200
Formulas and Flexibility
in Trade Negotiations
Sensitive Agricultural Products in the WTO's
Doha Agenda
Sebastien Jean
David Laborde
Will Martin
The World Bank
Development Research Group
Agriculture and Rural Development Team
February 2010
Policy Research Working Paper 5200
Abstract
Many trade negotiations involve large cuts in high objectives of policy makers in setting the prenegotiation
tariffs, with flexibilities allowing much smaller cuts for tariff. Applying this approach with detailed data allows
an agreed number of politicallysensitive products. The the authors to assess the implications of sensitiveproduct
effects of these flexibilities on market access opportunities provisions for average agricultural tariffs, economic
are difficult to predict, creating particular problems for welfare, and market access under the Doha negotiations.
developing countries in assessing whether to support The authors conclude that highesttariff rules are likely
a proposed agreement. Some widelyused ad hoc to seriously underestimate the impacts on average tariffs,
approaches to identifying likely sensitive productssuch and that treating even 2 percent of tariff lines as sensitive
as the highestboundtariff rulesuggest that the impacts is likely to have a sharply adverse impact on economic
of a limited number of such exceptions on average tariffs welfare. The impacts on market access are also adverse,
and on market access are likely to be minor. This paper but much smaller, perhaps reflecting the mercantilist
uses a rigorous specification based on the apparent focus of the negotiating process.
This papera product of the Agriculture and Rural Development Team, Development Research Groupis part of a larger
effort in the department to understand the implications of trade agreements for development. Policy Research Working
Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at wmartin1@worldbank.
org. This research was supported by the MultiDonor Trust Fund for Trade.
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Produced by the Research Support Team
Formulas and Flexibility in Trade Negotiations: Sensitive
Agricultural Products in the WTO's Doha Agenda
by
Sebastien Jean*, David Laborde** and Will Martin***
*INRA and CEPII, Paris; **IFPRI and ***World Bank, Washington DC. The views in this paper are those
of the authors alone and not necessarily those of the institutions with which they are affiliated.
Formulas and Flexibility in Trade Negotiations: Sensitive
Agricultural Products in the WTO's Doha Agenda
In recent years, a common feature of trade negotiations involving developing countries
has been the use of a formula approach to tariff cutting, coupled with provisions for
smaller, or zero, cuts in particular products. This approach follows a pattern observed in
earlier WTO Rounds (Martin and Winters 1996) and regional agreements (Olarreaga and
Soloaga 1998), where ambitious tariff reduction goals were combined with discretion for
particular, politicallysensitive, products. One difference is that the specific products to
be subjected to smaller cuts were typically directly negotiated in earlier agreements,
while the Doha agenda "modalities" for agriculture specify the share of products allowed
smaller cuts, leaving the choice of products to the discretion of the importer.
The approach being followed in the WTO agricultural negotiations (WTO 2004,
2008) specifies larger proportional cuts in higher bound tariffs but allows reduced cuts
for "sensitive" products selected by members. Earlier work suggests this approach may
make market access gains particularly susceptible to erosion through exclusion of a small
number of goods, particularly in the industrial countries where the variance of
agricultural tariffs is very high (Jean, Laborde and Martin 2006).
While discretion for smaller cuts on their own sensitive products is attractive to
individual policy makers, it raises difficult questions for policy makers and for analysts
making ex ante evaluations of proposed agreements. These problems are particularly
acute for policy makers from small developing countries. While large traders may have
the resources to estimate the direct impact of key partners' choices on their market access
reasonably well, small developing countries frequently have difficulty doing so. Both
groups remain vulnerable to importers changing their choice of products at the last
moment. Analysts attempting to provide policy makers with ex ante assessments of
proposed global agreements face a different and perhaps even more serious challenge
they need a consistent basis for assessing the use of these flexibilities in 153 WTO
member countries.
One widelyused approach to ex ante assessment is to assume that flexibility will
be used for the highest bound (Sharma 2006) or applied (WTO 2006) tariffs. These
approaches lead to a sharp conclusionthat the impacts of flexibilities on cuts in average
tariffs will be small. The highly variable, and frequently large, gaps between bound and
applied agricultural tariffs (Jean, Laborde and Martin 2006) raise important questions
would products with high bound tariffs really be selected if the bound rate remains above
the applied rate even after the bound rate is cut, and hence no reduction in applied rates is
required? Even high applied rates may not be subject to cuts if the binding overhang on
these products is sufficiently large. Further, these approaches ignore the importance of
the producthigh tariffs are frequently observed on very minor products.
To deal with these concerns, Jean, Laborde and Martin (2006) proposed a
minimizationoftariffrevenueloss rule that takes into account the size of the cut in
applied tariffs resulting from the formula and binding overhang, and the initial value of
imports. This approach leads to a strikingly different conclusionthat even a small
number of sensitive products can dramatically reduce the cuts in average agricultural
tariffs. In this situation, it seems particularly desirable to have an approach with stronger
behavioral foundations.
2
To estimate the effects of flexibilities requires a forecast of the products likely to
be chosen, and an estimate of the effects of these choices on efficiency and market
access. We first develop a simple, theoreticallywellgrounded, model of the preferences
of policy makers, and then use it to assess which agricultural products WTO members are
likely to treat as sensitive. Our approach focuses on policy choices within a single
country, building on the framework developed by Grossman and Helpman (1994) and
others in the politicaleconomy literature. It provides a muchneeded ex ante assessment
of the impact of policy choices on market access and welfare, and a basis for ex post
testing should the current negotiations be successfully completed.
An important question for current and future negotiations is whether the
combination of ambitious tariffcutting formulas with flexibilities allowing small cuts on
relatively high tariffs makes sense from economic or mercantilist perspectives. To shed
light on this issue, we use the AndersonNeary (2007) approach with the most
disaggregated data available at the international level to assess the implications of
flexibility for both welfare and market access.
We assume that the agricultural tariff prior to the negotiations results from
maximization of a government objective function along the lines of the Grossman
Helpman (1994) model. This approach seems appropriate for agricultural tariffs in the
current negotiations because they have not been effectively disciplined by multilateral
agreements (Hathaway and Ingco 1996). We consider a liberalization agreement
involving a tariffreduction formula together with flexibility for sensitive products, and
focus on the way policy makers use this flexibility at any given level of world prices. The
combined effect of the decisions by all 153 members leads to changes in domestic and
3
international prices that members must take into account in deciding whether to accept
the agreement.
Our first step is to develop an objective function for government policy making.
Then, in Section II, we use this function to assess the implications of changes in tariffs
under alternative assumptions about the structure of preferences. In Section III, we
discuss the data and tariffcutting formulas on which we base our analysis. Section IV
contains applications to realworld data designed to assess the likely outcomes for
average tariffs; to provide comparisons with earlier approaches; and to examine the
sensitivity of outcomes to different rules for sensitive products. In Section V, we examine
the implications of sensitive products for economic welfare in the country utilizing the
flexibility and for the market access opportunities of partner countries. Section VI
concludes.
I. Representing Governments' Objective Functions
We begin by specifying an objective function for policy makers that takes into account
the benefits to politicians from providing protection to particular sectors while
considering the costs to consumers and taxpayers of providing this protection. Our
politicaleconomy objective functionbased on Grossman and Helpman (1994, equation
5)is expressed as:
G (p, u ) =  z (p, u ) + z p ' (p  p*) + h' p (1)
Where z(p,u)=e(p,u)g(p) is the trade expenditure function, defined as the difference
between the consumer expenditure function e(p,u) defined over domestic prices, p and
the utility level of the representative household, u, and a net revenue function, g(p),
defined over domestic prices for given factor endowments; p* is the vector of world
4
prices for traded goods, so that (pp*) is a vector of specific tariff rates; zp = ep gp is a
vector of net imports; zp´(pp*) is tariff revenues, assumed to be redistributed to the
household; and the elements of h reflect the valuation by governments of changes in
domestic prices, over and above their impact on general economic welfare. We consider
only tariffs because domestic and export subsidies are dealt with under different "pillars"
of the negotiations.
Like Grossman and Helpman (1994, proposition 2) and virtually all subsequent
applications based on this model, we assume that importers and governments view import
prices as fixed, so that changing tariffs from their initial level involves a reduction in the
value of the government's objective function. For individual governments choosing their
own protection levels on individual commodities, this seems reasonable for relatively
homogenous agricultural products given that the estimated export supply elasticities for
homogenous goods are five times as high as for other products (Broda, Limao and
Weinstein 2008, p2033). It also seems consistent with the approach taken by agricultural
policy makers dealing with productspecific issues such as the "tariffication" of nontariff
barriers (Hathaway and Ingco 1996).
If we move beyond the GrossmanHelpman (1994) model, the h weights may also
reflect a number of politicaleconomy features identified by authors such as Anderson
and Hayami (1986), Lindert (1991), Olarreaga and Soloaga (1998), Cadot, de Melo and
Olarreaga (2004), and Dutt and Mitra (2010) that influence how much protection a
particular agricultural sector will receive. These include: (i) how effectively the sector is
organized; (ii) the impact of ownoutput prices on returns to specific factors in that
sector; (iii) adverse impacts on the costs of other politicallyinfluential groups of
5
protecting a particular sector; and (iv) the ratio of imports to domestic consumption that
determines the balance of benefits between tariff revenues and transfers to producers, and
(v) the degree of concentration in the sector. In contrast with the studies above, our
objective is not to explain the premium placed by policy makers on higher prices for
particular goods. Rather, we use the observed policy choices to infer the elements of h 
something that is feasible for highly disaggregated products.
Since we assume that the politicaleconomy objective function is being
maximized in the initial equilibrium, we can use the first order conditions to solve for h:
h = z pp (p 0  p*)
0
(2)
where zpp0 (p0p*) is the marginal welfare cost of tariff changes around (pp*), and the
superscript 0 refers to values at the initial equilibrium (since world prices are assumed to
be constant, p*0 = p*). The revealed value of h for product i clearly depends on the tariff
for that sector. However, hi depends also on the slope of the demand curve, zii, and the
crossprice effects with other goods subject to tariffs, zij. In addition, for any given import
demand elasticity, the value of hi increases with import volume. Note that hi for a good
with a zero tariff will be negative if there are positive tariffs on its substitutes and none
on any complements. Sectors that are organized will likely have positive values of hi
while unorganized sectors are expected to have negative values. Equations (1) and (2)
together show the strong link between our approach and the GrossmanHelpman (1994)
formulation. 1
At any point where producers and consumers are making optimizing decisions
relative to domestic prices, zppp =0. Equation (2) can thus be simplified to:
h = z pp p *
0
(2)
6
Equation (2) includes all prices, and must yield exactly the same estimates as (2). It
allows us to rewrite (1) in terms of potentially observable terms:
G =  z (p, u ) + z p (p  p*) + p*' z pp p
0
(1)
Equation (1) provides the basis for our subsequent analysis.
II. Implications of Tariff Changes for the Objective Function
A secondorder TaylorSeries expansion of equation (1) around the initial equilibrium
provides insights into the implications of tariff changes that change p relative to p*. We
begin by taking the first and second derivatives of (1) with respect to prices:
G 2G
= p*' z pp + (p  p*)' z pp
0
and = z pp + z ppp (p  p*) (3)
p p 2
For lack of information about the third derivatives of the trade expenditure function, we
assume that the trade expenditure function can be adequately represented by a function
such as the normalized quadratic introduced by Diewert and Ostensoe (1988) or the
symmetric normalized quadratic used by Kohli (1993) to model import demand. As noted
by these authors, these are flexible functional forms and hence can provide a second
order approximation at any point to any twicedifferentiable functional form, such as the
widelyused, but much less flexible, CES function. Given this assumption, the zppp term
in equation (3) can be dropped and the implications of deviations in tariffs from the
domestic politicaleconomy optimum can be analyzed using the TaylorSeries expansion:
G 1 2G 1
G = p + p' 2 p = p' z pp p (4)
p 2 p 2
Equation (4) is particularly simple because the initial equilibrium is an optimum from the
point of view of the government acting unilaterally. It contains none of the interactions
7
with existing distortions that complicate calculation of standard welfare effects (see
Martin 1997). The quadratic nature of equation (4) immediately reveals a problem with
the tariffrevenueloss rule of Jean, Laborde and Martin (2006)the impact of a
reduction in the required tariff cut on the value of the government's objective function
will depend not only on the size of the price increase allowed by sensitive product status,
but also on the size of the initial cut required by the formula.
Further insights into the effects of particular tariff changes can be obtained by
rearranging (4) into proportional change form:
s111 s112 ... s11n p1
^
s s2 22 p
G 1 ^
= [ p1
^ ^
p2 ... 0] 2 21 2 (4')
e 2 ... ...
where e is initial expenditure on all goods and services, including the nondistorted
numeraire, n; si is the share of expenditure on good i; ij is the elasticity of demand for
^
good i relative to the price of good j; and the vector p refers to proportional changes in
domestic prices. Where the bound tariff equals the applied, i = (1+ ) where ci is the

tariff cut required by the formula. The relationship is less direct, but still readily
computable, when bound tariffs exceed applied rates. We express G relative to e,
without loss of generality, because this allows us to use value shares, rather than gross
values, as weights on the elasticity matrix.
For simplicity and tractability, a productbyproduct analysis can prove useful.
This is possible based on (4´), since the impact of allowing sensitiveproduct treatment
for product i on the government's objective function can be computed as the difference
8
G
between the welfare loss with the formula applied to product i,  f (i ) and with
e
G
sensitiveproduct treatment, s (i ) . If we let pi represent the impact of the formula cut
^
e
on pi and the cut with flexibility be ( pi + ~i ) where ~i 0 is the increase in the price
^ p p
from the postformula level as a proportion of its initial domestic price, we obtain,
following the steps outlined in the Technical Appendix:
G G 1
 = si ~i[ii ~i + 2 ij p j ]
p p ^ (5)
e s (i ) e f (i )
2 j
A key insight from equation (5) not available from equation (4) is the potential
importance of the size of the formula cuts on other goods for the selection of good i.
Unfortunately, we do not have the matrix of own and crossprice elasticities for
over 5000 products included in equation (4'). If we use CES preferences to obtain local,
theoreticallyconsistent, estimates of these elasticities, the ownprice elasticities are given
by (1si). , where is the elasticity of substitution, and the crossprice elasticities, ij are
given by .sj. As shown in the Technical Appendix, equations (4') and (5) can then be
rewritten including crossprice effects as:
G 1 1
= s j p j pi si  p j =  .VAR( p )
^ ^ ^ ^ (4'')
e 2 j i 2
^
with VAR( p ) the weighted ( si ) variance of price changes pi ; and
^
G G 1
 = .si ~i[(1  si )( ~i + 2 pi ) + 2 s j p j ]
p p ^ ^ (5')
e s (i ) e f (i )
2 j i
Note that in this CES framework the choice of sensitive products is independent of .
Two features of equation (5') allow us to identify a potential simplified rule for
selecting individual tariff lines: (i) since dutiable agricultural goods usually represent a
9
small share of total expenditure, including goods that are not imported subject to tariffs,
s j p j is likely to be negligible in (5'); and (ii) with such a large number of sensitive
^
j i
products, it is likely that s j p j is very similar for almost all choices of potential
^
j i
sensitive products, leaving the ranking of products unchanged even if s j p j is
^
j i
nontrivial. With either or both of these assumptions, a simplified measure of the impact
of designating product i as sensitive is:
G G 1
  si (1  si )(( pi + ~i ) 2  pi )
^ p ^2 (6)
e s (i ) e f (i )
2
In the CES case, equation (6) provides a simple rule of thumb for selecting sensitive
products that depends only on observable information on the expenditure share of the
good in the presence of tariffs, si; the size of the price cut implied by the formula and any
in its price, i . The intuition of this measure is clear: it compares two trianglesformed
binding overhang; and the extent to which sensitive product selection allows a smaller cut
by multiplying the elasticity of import demand (1si) by a squared proportional change
in prices to measure the reduction in the loss of policy maker welfare when sensitive
products are allowed.
The logic of the simplification involved in moving from equation (5') to equation
(6) might also be used to justify a similar simplification of equation (5):
G G 1
  siii (( pi + ~i ) 2  pi )
^ p ^2 (7)
e s (i ) e f (i )
2
This approach follows Feenstra (1995) in using just the ownprice terms to assess
the implications of a tariff regime. It is particularly attractive since Kee, Nicita and
10
Olarreaga (2008) provide estimates of exactly the ownprice elasticities required for this
approach. 2
III. Data and Tariff Formulas
We use the MAcMapHS6 v1.1 database (Bouët et al. 2008) on applied protection in the
base year for the negotiations, 2001. This dataset includes key features such as the ad
valorem equivalents of specific tariffs; an assessment of the impact of tariffrate quotas
(TRQs) on many key commodities; tariff preferences, and import values. The analysis is
carried out at the finest level at which country classifications are internationally
compatible: the sixdigit level of the Harmonized System. 3 The protective effect of TRQs
is represented by using the inquota tariff when the quota is less than 90 percent filled;
the outofquota tariff when the quota is filled; and their average in between. Adjustments
were made for Korean corn and soybeans, where imports over high outofquota tariffs
appear to be very large, but the TRQs are, in fact, expanded to meet demand. A pre
experiment introduced reforms that will proceed irrespective of the Doha outcome,
including expansion of the European Union, the phasein of remaining agricultural
commitments by developing countries, 4 and reforms agreed by WTO accession countries.
The negotiations specify cuts in WTO bound tariffs, which are frequently well
above applied rates. This binding overhang means that reductions in bound tariffs will not
always bring about corresponding reductions in applied rates or increases in market
access. A detailed dataset on bound duties (see Bchir et al. 2006) conformable with the
MAcMapHS6 applied rate data was used to specify the cuts in bound rates. Applied rates
were reduced to the extent that the new bound rate declined below the initial applied rate.
11
The analysis uses the tariff cut proposal that has shaped the negotiationsthe
proposal by the G20 of a tiered formula with four bands and three inflexion points (G20
2005). For the industrial countries, this involves proportional cuts in bound tariffs that
increase through four tiers to reach 75 percent on tariffs above 75 percent. For developing
countries, the cuts rise to 40 percent on tariffs above 130 percent. Tariffs are capped at
100 percent for developed countries and 150 percent for developing countries. Least
developed countries are not required to undertake any reduction commitments.
Bound tariffs on sensitive products can be cut by one third or two thirds of the
formula cut, with increases in TRQs required to compensate trading partners for the
resulting loss of market access (WTO 2008). We assume that the combined effect of the
tariff cut and TRQ expansion for a sensitive product with a TRQ is onehalf the formula
cut.
IV. Experiments and Impacts on Average Tariffs
We used four different approaches to identifying sensitive products. The first was to
solve equation (1') using nonlinear integer programming. However, we encountered
multiple solutions using this nonlinear approach. We therefore turned to approaches
based on the secondorder approximations discussed in Section II. Our initial results were
obtained by solving equation (4'') using the SBB (Branch & Bound) GAMS® solver for
Mixed Integer Nonlinear Programming (MINLP) (see GAMS 2010) with up to 2 percent
of products allowed as sensitive. This was complemented by simple oneproductata
time selections using equation (6), and equation (7) with the ownprice elasticities of
Kee, Nicita and Olarreaga (2008).
12
The scenario against which we assess the impact of sensitive products applies the
formulas to all products, without exception ("Formula" column in Table 1). We first
compare these results against those using our three approaches to identifying two percent
of sensitive products. We then compare these results with those from the three ad hoc
approaches used in earlier studies "highest bound", "highest applied" and "tariff
losses". Next, we examine the potential sensitivity of our results to whether or not
sensitive products include "sin" products, which might have high tariffs to discourage
consumption rather than for politicaleconomy reasons. Then, we consider the sensitivity
of the results to the number of tariffs allowed sensitive treatment. Finally, we consider the
implications of an alternative approach of basing the share of sensitive products on the
percentage of imports, rather than a percentage of tariff lines.
Our results for the "Formula" scenario are given in the second column of Table 1.
Even though the formulas more than halve average bound tariffs worldwide, the
reductions in applied rates are smaller because of binding overhang. With no sensitive
products, the average tariff for nonLDC WTO members is cut by 6 percentage points,
from 14.6 percent to 8.6 percent (Table 1, column "Formula"). Among the main countries
shown in Table 1, only Canada, the EU, EFTA, Japan and South Korea display more than
a 5 percentage point cut in applied rates. Indeed liberalization appears to be
overwhelmingly concentrated in Japan, EFTA and Korea, with very limited liberalization
elsewhere. 5 For many countries, applied duties are hardly changed: 8 out of the 18
countries and groups shown in Table 1 experience a decline in applied duties of less than
two percentage points. The formula considered narrows the binding overhang in many
cases, without substantially changing applied rates.
13
Table 3 displays the products most frequently selected as sensitive by developed
and developing countries when 2 percent of sensitive products are selected
simultaneously using equation (4''). For comparison purposes, Appendix Table 1
compares these products with those selected using the highestaverage tariff rules, and
considers the coverage of agricultural imports for the different sets of products
considered. This comparison shows that selecting products based on the highest tariffs
leads to frequent inclusion of minor products, such as "foliage branches", "maize
stalks", and "garlic" in the industrial countries and "other cereals", with a share in
agricultural imports of only onehundredth of a percent in developing countries. A
striking difference between the lists selected using equation (2´) and the highesttariff
rules is in the share of agricultural imports covered. Our politicaleconomy approach
results in a list of mostcommon sensitive products covering 80 (63) percent of
agricultural imports into the industrial (developing) countries, while the highest bound
tariff rule leads to a list covering only 5 (7) percent.
The resulting impacts for countries' own weightedaverage tariffs are presented
in Table 1 (column "Sens 2"). Allowing 2 percent sensitive products, the cut in the
world wide average applied duty drops from 6 percentage points to 3.1. Four relatively
highlyprotected Harmonized System chapters, Meat and offal (02), Cereals (10), Fruits
(08) and Sugar (17), accounting for 27 percent of total imports, contribute 67 percent of
the tariff cut without exclusions, but 80 percent of the reduction in tariff cuts when
sensitive products are introduced. For developing countries, four chapters (01Meat, 10
Cereals, 12Oilseeds and 24Tobacco) cover 25 percent of total imports and contribute
51 percent of the basic cut but 64 percent of the reduction in the cut.
14
Using the simplified criterion in equation (6) ("Sens 2simple") changes the
aggregate results very little. This is reassuring, given that algorithms for simultaneous
product selection are unlikely to be available to policy makers. The results based on
equation (7) ("Sens 2elas") are somewhat higher than those for "Sens 2simple",
reflecting the fact that the elasticity criterion used in their selection does not enter the
calculation of the standard tradeweighted averages. 6
An important question is how our results compare with the adhoc alternatives
used in earlier policy analyses. Scenario "Sens 2highest bound" uses Sharma's (2006,
p5) ruleofthumb of selecting the products with the highest bound tariffs. This
approach turns out to yield dramatically lower estimates of the impact of sensitive
products on applied rates: the cut in the average applied tariff is found to decline by just
over one percentage point when sensitive products are selected this way. If products
with the highest applied tariffs ("Sens 2highest applied") are chosen as sensitive
(Martin and Wang 2004; WTO 2006), the impact on average tariffs is still much
smaller. These two rules select many minor products with high tariffs.
Column (8), "Sens 2tariff loss" selects sensitive products by minimizing tariff
revenue losses. At the aggregate level, the results using this criterion differ little from
those using our politicaleconomy approach. At a disaggregated level, we find that our
politicaleconomy criteria pick some productssuch as virgin olive oil for the
European Unionthat seem likely to be treated as exceptions, but are not identified
using the tariff revenue loss criterion.
As is clear from Table 3, some of the WTOagricultural products selected as
"sensitive" are "sin" tax commodities such as cigarettes or alcohol. If high duties on these
15
products are being used to raise revenues or to reduce externalities, countries may not
follow a politicaleconomy rule when choosing sensitive products. To guard against this,
"Sens 2sin" is derived using the same approach as "Sens 2" but excluding "sin"
commodities such as alcohol and tobacco from the sensitive product category. This
exclusion is found to lead to cuts in average tariffs that are similar to those without this
exclusion. The increase in the cut with this exclusion is from 4.3 to 4.5 percent in the
industrial countries, and from 1.2 to 1.6 percent in developing countries. This exclusion
does change the composition of the products selected. In developed countries,
preparations of meat and fish, and dairy products become more important, as do dairy
products, fruits, meats and fats in developing countries.
While our analysis so far has focused on allowing 2 percent of tariff lines, many
WTO members have sought much higher percentages of tariff lines. 7 We find that raising
the number of sensitive products to 4 percent ("Sens 4") using the political economy
criterion in equation (4'') has only a small impact, except in a few cases such as Japan and
EFTA. Overall, the extent of delivered liberalization is only slightly reduced because
sheltering just 2 percent of products is enough to greatly reduce the cut in average tariffs.
While the agricultural negotiations under the Doha Agenda have focused on
restricting the number of sensitive products, constraints on the value of trade have been
used in some other negotiations. 8 To shed light on the differences between constraints
based on trade and those based on tariff lines, we compare the results for "Sens 2" and
"Sens 4" with those where imports are constrained by import value"Sens 2trade" and
"Sens 4trade". This comparison shows considerable differences. The global reduction in
average tariffs is 4.8 percent under "Sens 2trade", as against 3.1 percent under "Sens 2".
16
As compared to the "Formula" scenario, allowing 2 percent of imports as sensitive
products based on trade causes the cut in world average tariffs to decline from 6.0 percent
to 4.8 percent, with limited reductions in the resulting tariff cuts in most cases, in contrast
with the dramatic and unpredictable reductions in disciplines associated with sensitive
product limits based on tariff lines.
Figure 1 illustrates the relationship between the number of sensitive products and
the final level of tariffs in a general way, by plotting the relationship between the number
of sensitive products allowed, and the average level of applied protection. When the
constraint is expressed in terms of number of products, the curve is indeed extremely
steep near the yaxis: a very small share of sensitive products is enough to sweep out a
significant part of the applied tariff cut. This is even clearer for developed countries than
for developing countries. When defined as an import share, in contrast, changes in the
number of sensitive products have a far less precipitous impact on tariff cuts. As far as
developed countries are concerned, allowing 5 percent of initial imports to be defined as
sensitive products reduces tariff cuts by approximately one third, 10 percent of imports
would reduce them by almost twothirds.
While trade is also an imperfect criterionsince highlyrestricted products are
likely to have small importsit seems clear that its deficiencies as a basis for specifying
sensitive products are less serious than those associated with the number of tariff lines.
There is an important underlying reason for this better performanceexternal trade
reflects the interests of the exporter rather than solely those of interest groups within the
importing country.
17
Table 2 presents results corresponding to those in Table 1 for the case of 2 percent
sensitive products but this time for the protection faced by each country. A sharp
difference between the results for protection applied and protection faced is evident for
developing countries. For most developing countries, allowing sensitive products reduces
the extent of required ownliberalization very little because the cuts in their own applied
rates in the absence of sensitive products are quite small. By contrast, allowing sensitive
products results in a very substantial reduction in market access gains. The average
reduction in tariffs facing developing countries declines by almost 3 percentage points
from 5.5 percent to 2.7 percent.
V. Implications for Welfare and Market Access
The average tariff measures reported in Tables 1 and 2 provide a broadand widely
understoodindication of the consequences of including flexibilities for the economic
welfare and for market access. However, it is well known that the weighted average tariff
is a flawed indicator of the efficiency or market access impacts of reform.
Anderson and Neary (2007) propose an integrating treatment of the problems of
aggregation and the implications of trade reforms for welfare and market access. Their
results provide a rigorous link between means and variances of tariff (specifically,
generalized means and generalized variances that reflect substitution relationships
between goods) and key policy outcomes including economic welfare and market access.
For the special model in which the expenditure function over all goods (domestic and
imported) takes the ConstantElasticity of Substitution form, and domestic and imported
goods are imperfect substitutes, the needed measures of the economywide generalized
mean and variance can be calculated easily.
18
Using these estimates of the generalized means and variances, we can assess the
implications of the flexibilities considered in this article for welfare in the importing
countries, and for the market access available to their partners. A key finding of
Anderson and Neary (2007) is that there are important differences in the impact of an
increase in the variability of tariffs on welfare and on market access. Increases in the
generalized variance of a tariff regime reduce welfare but will expand market access at a
constant generalized mean (Anderson and Neary, 2007, p192, equation 16 and p193,
equation 21). It seems likely that allowing sensitive products will increase the variance of
the trade regime. Questions for policy makers therefore arise. Will a policy of allowing
sensitive products have a less adverse impact on partners' market access than it has on
the welfare of the country using the flexibility? And what are the magnitudes of these
impacts?
A key issue is the impact of changes in the mean and variance of tariffs from
allowing sensitive products for welfare and for market access. Figure 2(a) shows that the
real income loss due to sensitive products is, in most countries, more a function of the
rise in the tariff variance than of the mean tariff. By contrast, Figure 2(b) shows that the
increased tariff variance associated with sensitive products substantially reduces the
marketclosing impacts of the increase in the mean. These findings are consistent with the
demonstration by Kee (2007) and Kee, Nicita and Olarreaga (2008) that increases in the
weighted variance raise the efficiencyoriented Trade Restrictiveness Index, while not
affecting the mercantilist measure of trade restrictiveness. They highlight a reason that
negotiators in a mercantilist forum like the WTO might choose deep tariffcutting
19
formulas combined with exceptionsthe damage to market access is less than the
damage to efficiency.
These results imply that it is important to look beyond average impacts when
analyzing the impact of free trade on efficiency and market access. Reductions in tariffs
resulting from the formula approach raise welfare both through the reduction in the
generalized mean tariff and through reductions in the generalized variance. These results
strongly reinforce the need to go beyond average impacts.
VI. Concluding Remarks
The impact of exceptions from tariffcutting formulas in the Doha negotiations has been a
major source of uncertainty and conflict. Some widelyused rules of thumb for their
selection suggest that their overall impacts would be minor. We derive approaches to
selection based on a politicaleconomy framework applicable at a fine level of
disaggregation, and estimate likely impacts on key outcomes from the negotiations.
We show that allowing even a limited number of products to be subjected to
smaller cuts is likely to substantially reduce the extent of trade liberalization in the
developed countries, while developing countries gain little in this mercantilist sense. In
contrast, the costs to exporters are shared, and developing countries see their market
access gains fall substantially with sensitive product exceptions.
We find that approaches to sensitiveproduct identification based only on the
height of the tariff greatly underestimate their impact. However, we identify a simple
approach using only readilyavailable information on the share of the product at domestic
prices, the depth of the formula cut and the "relief" provided by flexibility that generates
20
impacts consistent with our more complex models. The tariffrevenueloss criterion used
in our earlier work appears to track closely the overall impact of our full model results.
A problem for exporters associated with allowing a certain number of tariff lines
to be sensitive is that this criterion does not take into account the importance of these
tariff lines to the exporter. If we do this in a crude way by restricting the number of
products on the basis of their share in total imports, we find a dramatic reduction in the
damage to market access created by sensitive products.
Building on recent work by Anderson and Neary (2007), we show in addition
that, since these exceptions increase the variance of tariffs relative to the formula
outcome, their effects on economic welfare are much worse than their effects on market
access. In this sense, the combination of steeply progressive tariff formulas and
exceptions may be much more rational from a mercantilist point of view than when
examined from the perspective of economic welfare and development.
21
References
Anderson, K., Y. Hayami and Others 1986. The Political Economy of Agricultural
Protection: East Asia in International Perspective, Boston, London and Sydney:
Allen and Unwin.
Anderson, J.E. and J.P. Neary 2003. "The Mercantilist Index of Trade Policy",
International Economic Review 44(2):627649.
Anderson, J.E. and J.P Neary 2007. "Welfare Versus Market Access: the Implications of
Tariff Structure for Tariff Reform." Journal of International Economics 71:187
205.
Bchir, M., S. Jean and D. Laborde 2006. "Binding Overhang and Tariffcutting
Formulas." Review of World Economics 142(2):207232.
Bouët, A., Y. Decreux, L. Fontagné, S. Jean and D. Laborde 2008. "Assessing Applied
Protection at the World Level." Review of International Economics, 16(5):85063.
Broda, C., N. Limão and D. Weinstein 2008, "Optimal Tariffs and Market Power: The
Evidence." American Economic Review 98(5):203265
Broda, C. and D. Weinstein 2006. "Globalization and the Gains from Variety" Quarterly
Journal of Economics 121(2):54185.
Cadot, O., J. de Melo and M. Olarreaga 2004. "Lobbying, counterlobbying and the
structure of tariff protection in rich and poor countries" World Bank Economic
Review 18(3):34566.
Diewert, E. and L. Ostensoe 1988. "Flexible Functional Forms for Profit Functions and
Global Curvature Conditions" in Barnett, W., E. Berndt and H. White eds.
Dynamic Econometric Modeling, Cambridge University Press, Cambridge.
Dutt, P. and Mitra, D. 2010. "Explaining Agricultural Distortion Patterns: The Roles of
Ideology, Inequality, Lobbying and Public Finance" in Anderson, K. (ed.), The
Political Economy of Agricultural Price Distortions, Cambridge University Press,
Cambridge. forthcoming Fall 2010
Feenstra, R. 1995. "Estimating the Effects of Trade Policy.", in G. Grossman and K.
Rogoff eds. Handbook of International Economics, Vol. 3, Amsterdam, Elsevier.
G20 2005. "G20 Proposal on Market Access." World Trade Organization, 12 October.
GAMS 2010. "SBB Solver Manual", Available at www.gams.com/solvers/solvers.htm
Grossman, G. and E. Helpman 1994. "Protection for Sale." American Economic Review
84(4): 83350 September.
Hathaway, D. and M. Ingco 1996. "Agricultural Liberalization and the Uruguay Round"
in Martin, W. and L.A. Winters, eds. The Uruguay Round and the Developing
Countries, Cambridge and New York: Cambridge University Press.
Hummels, D. 2001. Toward a Geography of Trade Costs, Unprocessed, Purdue
University.
Jean, S., Laborde, D. and Martin, W. 2006. "Consequences of Alternative Formulas for
Agricultural Tariff Cuts." in Anderson, K. and W. Martin eds. Agricultural Trade
Reform and the Doha Development Agenda. New York and Washington DC:
PalgraveMacmillan and the World Bank.
Kee, H. 2007. "Review of Measuring the Restrictiveness of International Trade Policy"
Journal of International Economics 73:43441.
22
Kee, H. A. Nicita, and M. Olarreaga 2008. `Import Demand Elasticities and Trade
Distortions' Review of Economics and Statistics 90(4):66682.
Kohli, U. 1993. "A symmetric normalized quadratic GNP function and the US demand
for imports and supply of exports." International Economic Review 34(1):24355.
Lindert, P. 1991. "Historical Patterns in Agricultural Policy." in Timmer, C.P. ed.
Agriculture and the State. Ithaca NY:Cornell University Press.
Martin, W. 1997. "Measuring welfare changes with distortions" in Francois, J. and K.
Reiner teds. Applied Methods for Trade Policy Analysis. Cambridge University
Press, Cambridge.
Martin, W. and Z. Wang 2004. "Improving Market Access in Agriculture." mimeo,
World Bank, Washington DC.
Martin, W. and L. A. Winters 1996. The Uruguay Round and the Developing Countries.
Cambridge: Cambridge University Press.
Olarreaga, M. and I. Soloaga 1998. "Endogenous tariff formation: the case of Mercosur."
World Bank Economic Review 12(2):297320.
Sharma, R. 2006. "Assessment of the Doha Round Agricultural Tariff Cutting
Formulae." Downloadable from http://www.faologe.ch/Tariffcuts%20
%20Sharma%20%20Jan%2006b.pdf
WTO 2004. "Doha Work Program: Decision Adopted by the General Council on 1
August 2004." WT/L/579, World Trade Organization, Geneva. 2 August.
WTO 2006. "Agriculture Simulations." Excel file, World Trade Organization, Geneva.
WTO 2008. "Revised Draft Modalities for Agriculture." TN/AG/W/4/Rev.4, World
Trade Organization, Geneva, 6 December.
23
Table 1. Implications of Sensitive Products for Reductions in Countries' Average Applied Tariffs
Base Formula Sens 2 Sens 2 Sens 2elas Sens 2highest Sens2highest Sens 2 Sens 2sin Sens 4 Sens 2
simple bound applied tariff losses trade
Country: % percentage point cut
Industrial 14.9 8.5 4.3 4.4 4.7 7.4 7.2 4.3 4.5 3.8 6.8
Australia 3.1 1.0 0.5 0.5 0.6 1.0 1.0 0.5 0.8 0.5 0.8
Canada 9.8 5.0 1.5 1.5 1.9 4.8 4.8 1.5 1.5 1.0 3.8
EFTA 28.9 14.2 7.6 7.5 8.9 14.1 14.1 7.5 7.8 6.1 11.0
EU 13.4 7.5 4.4 4.4 4.4 6.4 5.9 4.4 4.4 4.0 6.3
Japan 35.6 22.4 11.2 11.3 12.2 19.1 19.1 11.0 11.2 9.9 18.0
USA 2.7 0.9 0.4 0.4 0.5 0.9 0.9 0.4 0.4 0.3 0.6
Developing 14.2 2.5 1.2 1.3 1.4 2.1 2.0 1.2 1.6 1.1 1.8
ASEAN 8.9 2.3 0.8 0.8 1.1 1.2 1.0 0.8 2.2 0.8 1.9
China 10.2 2.7 1.8 1.8 1.9 2.2 2.6 1.8 1.8 1.7 2.5
India 55.4 3.6 1.9 1.9 1.9 3.6 3.4 1.9 2.0 1.8 3.2
Korea 27.7 10.4 4.2 4.6 5.7 8.6 8.9 4.2 4.2 3.6 5.0
Maghreb 19.0 3.3 1.7 1.7 1.8 3.3 2.8 1.7 2.2 1.6 2.8
Mercosur 12.8 0.2 0.0 0.0 0.0 0.1 0.1 0.0 0.1 0.0 0.1
Mexico 9.5 0.9 0.2 0.2 0.2 0.9 0.9 0.2 0.3 0.2 0.8
Other SSA 25.3 2.0 0.9 0.9 1.0 2.0 1.1 0.9 1.5 0.8 1.9
Pakistan 31.3 0.1 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0
SACU 12.6 0.6 0.3 0.3 0.2 0.6 0.5 0.3 0.3 0.3 0.4
Turkey 14.1 1.1 0.5 0.4 0.6 1.1 1.1 0.4 0.5 0.4 0.6
ROW 10.3 1.8 1.0 1.0 1.0 1.4 1.3 1.0 1.4 0.9 1.5
NonLDC
WTO 14.6 6.0 3.1 3.1 3.3 5.2 5.0 3.0 3.2 2.7 4.8
Notes: Numbers in first column are weighted average agricultural tariffs in 2001 for nonLDC WTO members, adjusted for agreed reductions. Numbers in
subsequent columns are reductions in percentage points. Column headers name the scenarios: Base: 2001 applied tariffs; Formula: Applies the G20's tiered
formula (TF), without sensitive products (SPs); Sens 2: TF with 2% SPs selected according to eq. (4''); Sens 2simple: TF with 2% SPs, selected using eq. (6);
Sens 2elas: TF with 2% SPs, selected using eq. (7); Sens 2highest bound: TF with 2% SPs, selected by highest bound rates; Sens 2highest applied: TF with
2% SPs selected by highest applied rates; Sens 2tariff losses:TF with 2% SPs selected to minimize tariff loss; Sens 2sin: Sens 2 TF with 4% SPs selected using
eq. (4''); Sens 2trade: TF with 2% SPs , selected using eq. (4'').
24
Table 2. Implications of Sensitive Products for Reductions in Average Tariffs Faced
Base Formula Sens 2 Sens 2simple Sens 2elas Sens 2highest Sens2highest Sens 2tariff
bound applied losses
Country: % percentage point cut
Industrial ctries 15.3 6.5 3.4 3.3 3.7 5.8 5.8 3.3
Australia 19.1 9.6 4.4 4.4 4.9 8.3 8.5 4.2
Canada 9.6 4.5 2.0 2.0 2.2 4.4 4.5 2.0
EFTA 15.6 6.3 4.7 3.5 4.3 6.1 5.6 3.9
European Union 15.7 5.6 3.3 3.3 3.6 5.4 5.3 3.2
Japan 10.5 2.4 1.6 1.6 1.8 2.3 2.3 1.6
USA 16.3 7.5 3.6 3.6 4.0 6.3 6.3 3.6
Developing ctries 13.8 5.5 2.7 2.8 2.9 4.4 4.1 2.7
ASEAN 20.6 5.9 3.0 3.0 3.3 4.5 4.2 2.8
China 15.6 8.2 3.5 4.0 4.1 5.5 5.4 3.8
India 9.5 3.3 1.6 1.7 1.8 2.9 2.2 1.6
Korea 16.1 6.6 5.0 4.7 5.1 5.6 5.6 4.1
Maghreb 14.0 5.5 2.9 4.7 3.0 5.3 5.1 4.5
Mercosur 13.5 5.0 2.4 2.4 2.5 4.3 3.7 2.4
Mexico 3.8 1.6 0.8 1.0 1.1 1.5 1.4 0.8
Other SSA 10.5 5.2 2.3 2.3 2.3 3.3 2.8 2.2
Pakistan 14.3 6.0 3.2 3.2 3.7 5.8 4.0 3.1
SACU 18.1 7.2 4.5 4.4 4.3 6.9 6.3 4.4
Turkey 9.7 3.1 1.5 1.8 1.6 2.1 2.0 1.8
ROW 12.3 5.5 3.0 2.9 3.0 4.8 4.7 2.9
NonLDC WTO 14.6 6.0 3.1 3.1 3.3 5.2 5.0 3.0
Notes: Numbers in first column are weighted average agricultural tariffs in 2001 faced by WTO members on WTO markets, adjusted for agreed reductions.
Numbers in subsequent columns are reductions in percentage points. Column headers name the scenarios: Base: 2001 applied tariffs; Formula: Applies the
G20's tiered formula (TF), without sensitive products (SPs); Sens 2: TF with 2% SPs selected according to eq. (4''); Sens 2simple: TF with 2% SPs, selected
using eq. (6); Sens 2elas: TF with 2% SPs, selected using eq. (7); Sens 2highest bound: TF with 2% SPs, selected by highest bound rates; Sens 2highest
applied: TF with 2% SPs selected by highest applied rates; Sens 2tariff losses:TF with 2% SPs selected to minimize tariff loss; Sens 2sin: Sens 2 TF with 4%
SPs selected using eq. (4''); Sens 2trade: TF with 2% SPs , selected using eq. (4'')..
25
Table 3. Products Most Frequently Selected as "Sensitive"
Industrial Countries
1 0201 30 Fresh or chilled bovine meat, boneless
2 0202 30 Frozen, boneless meat of bovine animals
3 0207 14 Frozen cuts and edible offal of fowls of the species Gallus domesticus
4 0406 90 Cheese
5 0603 10 Fresh cut flowers and flower buds, for bouquets or ornamental purposes
6 0702 00 Tomatoes, fresh or chilled
7 1001 90 Wheat and meslin (excl. durum wheat)
8 1701 11 Raw cane sugar (excl. added flavoring or coloring)
9 2106 90 Food preparations, n.e.s.
10 2202 90 Nonalcoholic beverages (excl. water, fruit or vegetable juices and milk)
11 2204 29 Grape juice (including grape must)
12 2402 20 Cigarettes containing tobacco
Developing Countries
1 2402 20 Cigarettes containing tobacco
2 2208 30 Whiskies
3 2203 00 Beer made from malt
4 1701 99 Cane or beet sugar
5 2204 21 Wine of fresh grapes, incl. fortified wines in bottles
6 2208 70 Liqueurs and cordials
7 2208 90 Ethyl alcohol < 80% vol, not denatured; spirits and other spirituous beverages
8 0207 14 Frozen cuts and edible offal of fowls of the species Gallus domesticus
9 2403 10 Smoking tobacco
10 2106 90 Food preparations, n.e.s.
11 2208 60 Grape juice
12 1006 30 Semimilled or wholly milled rice, whether or not polished or glazed
13 1701 11 Raw cane sugar
14 1806 31 Chocolate and other preparations containing cocoa, in blocks, slabs or bars of <= 2 kg,
15 1806 90 Chocolate and other preparations containing cocoa
Note: fifteen products are included in the list for developing countries because the last four products were
selected the same number of times.
26
Figure 1. Average applied tariffs resulting from the application of the tiered
formula, depending on the criterion and threshold used to define sensitive products
14%
Developing countries, constraint in % of lines
13%
Developing countries, constraint in % of trade
Applied tariffs  Trade weighted average
12%
11%
Developed countries, constraint in % of lines
10%
9%
8%
Developed countries, constraint in % of trade
7%
6%
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
Share of sensitive products
Note: This graph plots the average applied protection level for nonLDC WTO members, once the
G20 formula is applied. The share of sensitive products is reported on the xaxis. It is alternatively
defined as a share in the number of agricultural products, or as a share in imports. Note that even
sensitive products experience a cut in tariffs, although it is smaller than under the formula, so that
the curves do not converge towards initial protection levels.
27
Figure 2. Welfare and market access losses from changes in generalized mean and variance of tariffs, 2% sensitive products
a. Welfare losses b. Market access losses
Australia Australia
Canada Canada
Without variance
European Union Without variance European Union effect
effect Iceland
Iceland
Japan
Japan
Switzerland
Switzerland Variance effect
USA Variance effect
USA
China
China
India
India
Indonesia
Indonesia
Israel
Israel
South Korea
South Korea
Morocco
Morocco
South Africa
South Africa
0.25% 0.15% 0.05% 0.05% 0.15% 0.25% 0.35% 0.45%
0.00% 0.05% 0.10% 0.15% 0.20%
Note: Graphs are based on Anderson and Neary's (2007) equation 16 for social welfare and equation 21 for market access. Since generalized moments are equal to
trade weighted ones, the welfare losses are proportional to income, and the market access impacts are proportional to total imports.
28
Appendix Table 1. Products most commonly selected using political economy and highesttariff rules
With Political Economy criterion Rank With highest Bound tariff Rank With highest applied tariff Rank
Industrial countries 020130 FRESH BOVINE MEAT BONELESS 1 020230 BONELESS FROZEN MEAT OF BOVINE AN 1 020230 BONELESS FROZEN MEAT OF BOVINE ANIMALS 1
Industrial countries 020230 BONELESS FROZEN MEAT OF BOVINE AN 2 040590 FATS AND OILS DERIVED FROM MILK 2 100630 SEMI MILLED OR WHOLLY MILLED RICE 2
Industrial countries 020714 FROZEN CUTS & EDIBLE OFFAL OF FOWLS 3 060491 FOLIAGE BRANCHES & PARTS OF PLANTS 3 240399 CHEWING TOBACCO SNUFF 3
Industrial countries 040690 CHEESE EXCL. FRESH CHEESE 4 070320 GARLIC FRESH OR CHILLED 4 040590 FATS AND OILS DERIVED FROM MILK AND DEH 4
Industrial countries 060310 FRESH CUT FLOWERS AND FLOWER BUDS 5 100620 HUSKED OR BROWN RICE 5 060491 FOLIAGE BRANCHES AND OTHER PARTS OF PLA 5
Industrial countries 070200 TOMATOES FRESH OR CHILLED 6 100630 SEMI MILLED OR WHOLLY MILLED RICE 6 070320 GARLIC FRESH OR CHILLED 6
Industrial countries 100190 WHEAT AND MESLIN 7 100640 BROKEN RICE 7 100620 HUSKED OR BROWN RICE 7
Industrial countries 170111 RAW CANE SUGAR 8 170111 RAW CANE SUGAR EXCL. ADDED FLAVOUR 8 100640 BROKEN RICE 8
Industrial countries 210690 FOOD PREPARATIONS N.E.S. 9 230890 MAIZE STALKS MAIZE LEAVES 9 170111 RAW CANE SUGAR EXCL. ADDED FLAVOURING 9
Industrial countries 220290 NON ALCOHOLIC BEVERAGES 10 240220 CIGARETTES CONTAINING TOBACCO 10 170199 CANE OR BEET SUGAR 10
Industrial countries 220429 WINE OF FRESH GRAPES 11 240399 CHEWING TOBACCO SNUFF 11 220830 WHISKIES 11
Industrial countries 240220 CIGARETTES CONTAINING TOBACCO 12 010111 PURE BRED BREEDING HORSES 12 220840 RUM AND TAFFIA 12
Industrial countries 020319 FRESH OR CHILLED MEAT OF SWINE 13 020130 FRESH OR CHILLED BOVINE MEAT BONEL 13 220850 GIN AND GENEVA 13
Industrial countries 020329 FROZEN MEAT OF SWINE EXCL. CARC 14 020319 FRESH OR CHILLED MEAT OF SWINE 14 220860 VODKA 14
Industrial countries 020713 FRESH OR CHILLED CUTS & EDIBLE OFFAL 15 020322 FROZEN HAMS 220870 LIQUEURS AND CORDIALS 15
% of Agricultural imports 80 5 8
Developing countries 240220 CIGARETTES CONTAINING TOBACCO 1 220710 UNDENATURED ETHYL ALCOHOL 1 240220 CIGARETTES CONTAINING TOBACCO 1
Developing countries 220830 WHISKIES 2 220300 BEER MADE FROM MALT 2 220300 BEER MADE FROM MALT 2
Developing countries 220300 BEER MADE FROM MALT 3 220860 VODKA 3 220830 WHISKIES 3
Developing countries 170199 CANE OR BEET SUGAR 4 240110 TOBACCO NOT STEMMED OR STRIPPED 4 240310 SMOKING TOBACCO 4
Developing countries 220421 WINE OF FRESH GRAPES INCL. FORT 5 240120 TOBACCO PARTLY OR WHOLLY STEMMED 5 220890 ETHYL ALCOHOL 5
Developing countries 220870 LIQUEURS AND CORDIALS 6 240220 CIGARETTES CONTAINING TOBACCO 6 220850 GIN AND GENEVA 6
Developing countries 220890 ETHYL ALCOHOL 7 020230 BONELESS FROZEN MEAT OF BOVINE AN 7 240399 CHEWING TOBACCO SNUFF A 7
Developing countries 020714 FROZEN CUTS AND EDIBLE OFFAL OF FOWL 8 020629 FROZEN EDIBLE BOVINE OFFAL 8 220429 WINE OF FRESH GRAPES INCL. FORTIFIED WI 8
Developing countries 220860 VODKA 9 040690 CHEESE EXCL. FRESH CHEESE 9 220710 UNDENATURED ETHYL ALCOHOL 9
Developing countries 240310 SMOKING TOBACCO 10 100300 BARLEY 10 220820 SPIRITS OBTAINED BY DISTILLING GRAPE WIN 10
Developing countries 210690 FOOD PREPARATIONS N.E.S. 11 100620 HUSKED OR BROWN RICE 11 220860 VODKA 11
Developing countries 100630 SEMI MILLED OR WHOLLY MILLED RICE 12 100630 SEMI MILLED OR WHOLLY MILLED RICE 12 220870 LIQUEURS AND CORDIALS 12
Developing countries 170111 RAW CANE SUGAR 13 100890 CEREALS EXCL. WHEAT AND MESLIN 13 240120 TOBACCO PARTLY OR WHOLLY STEMMED OR STR 13
Developing countries 180631 CHOCOLATE AND OTHER PREPARATIONS 14 120220 SHELLED GROUND NUTS 14 170199 CANE OR BEET SUGAR AND CHEMICALLY PURE S 14
Developing countries 180690 CHOCOLATE AND OTHER PREPARATIONS 15 020322 FROZEN HAMS SHOULDERS AND CUTS 15 220421 WINE OF FRESH GRAPES INCL. FORTIFIED WI 15
% of Agricultural imports 63 7 22
Note: A list of up to 2% of agricultural tariff lines is selected for each country using equation (6), and the frequency table is generated from these lists. Shares of
agricultural imports are calculated for industrial and developing country groups by expressing the value of imports sheltered by sensitive product treatment
relative to total imports for the group.
29
Technical Appendix
This appendix sets out the intermediate steps in the derivation of equations (4''), (5) and (5').
Our starting point is equation (4'), describing the change in the politicaleconomy objective
function resulting from price changes:
s111 s112 ... s11n p1 ^
s s p
G 1 ^
= [ p1 p2 ... 0] 2 21 2 22
^ ^ 2 (4')
e 2 ... ...
With pi being the relative change in domestic price of good i, ij the compensated cross price
^
elasticity of good i relative to the price of j ( ii is the compensated own price elasticity of good
i), and si the share of good i in total expenditure.
Moreover we assume that:
^
The price change is pi if product i is not sensitive, and as such subject to the full formula cut,
and ( p + ~ ) if product i is sensitive, and as such subject to a smaller cut ( p p + ~ 0 ).
^ p
i i
^i
^ p
i i
Without any flexibility for sensitive products, Equation (4´) could be rewritten as:
G 1
= p j s j pi ji
^ ^
e 2 j i
30
A. From Equation (4´) to Equation (5):
Looking at the change in policy makers' objective function resulting from using flexibility to
reduce the cut in tariffs, we have:
G G
= ( p j + ~ j )s j ( pk + ~k ) jk  p j s j pk jk
1 1
 ^ p ^ p ^ ^
e p+ ~
^ p
e p 2 j
^ k 2 j k
( p j + ~ j )s j ~k jk + 2 ( p j + ~ j )s j pk jk  2 p j s j pk jk
1 1 1
= ^ p p ^ p ^ ^ ^
2 j k j k j k
1 1 1 1 1
= p j s j pk jk + 2 ~ j s j pk jk + 2 ~ j s j ~k jk + 2 p j s j ~k jk  2 p j s j pk jk
2 j
^
k
^
j
p
k
^
j
p
k
p
j
^
k
p
j
^
k
^
1 1 1
= ~ j s j pk jk + 2 ~ j s j ~k jk + 2 p j s j ~k jk
2 j
p
k
^
j
p
k
p
j
^
k
p
Assuming that only product i, is sensitive, we have ~k = 0, k i and
p
G G 1 ~
pi si pk ik + pi si pi ii + pi p j s j ji
 = ^ ~ ~ ~ ^
e s (i )
e f (i )
2 k j
1~
pi pk si ik + p j s j ji + ~i si ii
^ = ^ p
2 k j
Due to the symmetry of the second order derivatives of the expenditure function (Clairaut's /
Schwarz's theorem), s j ji = siij . Substituting in the second term of the previous expression:
G G 1~
 = pi si pkik + p j siij + ~i siii
^ ^ p
e s (i ) e f (i )
2 k j
(5)
1
= si ~i ii ~i + 2ij p j
p p ^
2 j
B. From Equation (5) to Equation (5'):
Now, under the CES assumption: ij = s j , j i and ii = (1  si ) with the elasticity of
substitution.
31
G G 1 ~
 = si pi  (1  si ) ~i + 2 p j s j  2 pi si  2(1  si ) pi
p ^ ^ ^
e s (i ) e f (i )
2 j
1
= si ~i  (1  si ) ~i  2 pi + 2 s j p j
p p ^ ^
2 j
1
= si ~i  (1  si ) ~i  2( pi + si pi ) + 2 s j p j
p p ^ ^ ^
2 j i
1
= si ~i[(1  si )( ~i + 2 pi ) + 2 s j p j ]
p p ^ ^
2 j i
C. From Equation (4') to Equation (4''):
G 1
= p j s j pi ji
^ ^
e 2 j i
= p j s j pi si  p j (1  s j )
1
^ ^ ^
2 j i j
1
= p j s j pi si + p j s j  p j
^ ^ ^ ^
2 j i j
1
= p j s j pi si  p j
^ ^ ^
2 j i
Let's define p as the average price change, i.e. p = pi si . Then
^
i
G 1
e
= sj pj p  pj
2 j
^ ^ ( )
1 2
= p  sj pj
^ 2
2 j
1
=  VAR( p )
^
2
32
Endnotes
= 
1
Using the model and notation of Grossman and Helpman (1994, equation 15), hi is given by
1 ( )
where W is economic welfare and mj´ is the slope of the import demand function. From their equation (5), hj
equals where a is the value placed by policy makers on general economic welfare relative to political
contributions and Cj is the contribution schedule of sector i.
2
Alternative estimates of import elasticities of substitution provided by Hummels (2001) and by Broda and
Weinstein (2006) would allow us to take into account crossprice effects, but these studies focus on substitution
between products from different suppliers, rather than on the responses of aggregate imports which are our focus
in this paper.
3
In an earlier assessment of the impact of flexibilities, Martin and Wang (2004) found little difference between
results obtained using tariffline level and sixdigit data for the impacts on overall protection considered in this
paper. While policy makers undoubtedly view some products at finer levels as strongly differentiated, many
countries have notified their tariffs to the WTO at the sixdigit level, and few have notified beyond the eight
digit level.
4
Developing countries had 10 years from 1994 to implement their Uruguay Round commitments, as did
developed countries for a few products.
5
Assessment of tariffcutting formulae is complicated in the case of Japan and Korea by the existence of large
tariffrate quotas with prohibitive outofquota tariffs, the ad valoremequivalent of which is difficult to gauge.
Assessments based on tariffs and on observed price differentials were used to compute meaningful ad valorem
tariff equivalents for rice in Japan and for rice and corn in Korea.
6
As well as using the conventional tradeweighted averages, we estimated the Trade Restrictiveness Index (TRI)
and the Mercantilist Trade Restrictiveness Index (Anderson and Neary, 2003) using the approach of Kee, Nicita
and Olarreaga (2008). In the TRI case, the estimated cuts in tariffs are slightly larger when the elasticitybased
selection criterion is used. In most cases, the MTRI is cut by slightly less when using this criterion. This is
related to the correlation between the aggregator used and components of our selection criterion (trade value,
elasticity, and the square of the tariff). Our key resultthat the size of the cut in tariffs is greatly reduced when a
theoreticallyconsistent approach to product selection is usedproved robust to the choice of aggregator.
7
While 2 and 4 percent of tariff lines have been the most widely discussed proposals for sensitive products, the
EU earlier proposed allowing 8 percent of tariff lines in the industrial countries. WTO (2008) would allow
developing countries onethird more than the industrial countries. In addition, many developing countries have
sought flexibility for an additional 20 percent of "Special Products" subject also to criteria such as food security
and livelihood security.
8
The "substantially all trade" criterion for Free Trade Areas under GATT Article XXIV is frequently interpreted
as limiting exceptions under these agreements to no more than 10 percent of trade. The Doha negotiations on
nonagricultural products restrict flexibilities using both trade and tariff lines.
33