The paper contributes to the discussion about the revenue implications of trade reform by assessing the approximate fiscal revenue impact of different liberalization formulae under consideration in multilateral trade negotiations for a group of low- and middle-income countries. The study applies a linear optimization framework to data for bound tariffs, applied tariffs, and imports at the HS-6 digit level for 58 developing countries, and simulates results for different sets of import demand elasticities and developing country "flexibilities." While only a small number of countries face a significant impact, results point toward the need for complementary fiscal measures in the countries most affected by revenue loss.


The paper contributes to the discussion about the revenue implications of trade reform by assessing the approximate fiscal revenue impact of different liberalization formulae under consideration in multilateral trade negotiations for a group of low- and middle-income countries. The study applies a linear optimization framework to data for bound tariffs, applied tariffs, and imports at the HS-6 digit level for 58 developing countries, and simulates results for different sets of import demand elasticities and developing country "flexibilities." While only a small number of countries face a significant impact, results point toward the need for complementary fiscal measures in the countries most affected by revenue loss.

I. Introduction

Trade liberalization is associated with higher economic growth (Berg and Krueger, 2003).2 Indeed, since the 1980s, globalizers among developing countries achieved higher growth rates than non-globalizers (Dollar and Kraay, 2002). However, although the evidence of a close correlation between openness and growth is strong, the causality remains controversial due to measurement difficulties, problems of endogeneity, and the correlation between openness and other variables such as the quality of institutions and growth-supporting infrastructure.3

Reluctance toward trade liberalization has also been founded on shorter-term concerns about the impact of trade liberalization on fiscal revenue (Bevan, 1995; Blejer and Cheasty, 1990), in particular in countries that rely heavily on customs duties and other trade taxes.4 Accordingly, the potential fiscal implications of trade liberalization have been the subject of discussion in the context of the World Trade Organization (WTO) Doha Round,5 and featured prominently in other negotiations, such as those between the European Union and developing countries on Economic Partnership Agreements (EPAs).

The impact of tariff cuts on a particular country is largely an empirical issue, as it depends inter alia on the initial trade value and tariff level; the size and mode of the tariff cut; and import demand and supply elasticities. While the effects of trade liberalization may include a decline in revenue from trade taxes, such an outcome is not unavoidable. Studies indicate that the net effect of trade liberalization on revenue, including second-round effects, could be positive,6 since (i) a reduction in tariffs could lead to higher import volumes, as a result of both income and substitution effects;7 (ii) demand could shift to items with higher tariff rates as a result of an income effect; (iii) a depreciation of the exchange rate following trade liberalization could raise the value of imports and tariff revenues in local currency; and, (iv) over the longer term, revenue would be expected to increase as a result of higher economic growth, normally associated with trade liberalization.

Moreover, beyond the reduction in tariffs, the liberalization of the trade regime can involve a variety of measures, some of which would be revenue neutral or even serve to raise revenue (Ebrill, 1999). In particular, a conversion of non-tariff barriers (NTB) such as quotas, bans, and import licenses into tariffs would generate additional revenue, and administrative reforms could entail efficiency gains in customs administration. However, as the occurrence of NTBs has declined markedly, there is now less leeway to replace NTBs with customs duties in order to increase revenue from trade taxes (Table 1). The impact of trade liberalization on tariff revenue is also determined by the extent of exemptions and preferences, and tariff revenue would increase to the extent that exemptions are reduced or abolished. Incentives to smuggle or misrecord would be reduced if tariffs were lowered or consolidated in the interest of greater transparency, thereby improving compliance and broadening the tax base (Fisman and Wei, 2004; Greenaway and Milner, 1991).

Table 1.

Tariff Lines Subject to Non-tariff Barriers in Developing Countries, 1989–2000 1/

(In percent of total tariff lines)

article image
Source: World Bank (2004).

Regional averages of share of tariff lines subject to core NTBs, including quantitative restrictions, price administration, and monopolistic trading channels.

In the framework of multilateral trade liberalization, the difference between bound tariff rates8 (the subject of WTO negotiations) and applied statutory (most favored nation—MFN) tariff rates is also critical: in cases where applied rates are significantly lower than bound rates, the latter can be lowered in the context of trade liberalization agreements without a significant impact on revenue.9 However, in the event that trade liberalization took place in a bilateral/regional rather than in a multilateral (MFN) context, revenue could be affected negatively if bilateral or regional trade agreements diverted imports from dutiable to preferential sources, although the net effect of this would require detailed analysis. Accordingly, the net impact of trade liberalization measures depends critically on a range of assumptions. While first-round effects are relatively easy to quantify, both the timing and the strength of second-round effects, reflecting behavioral responses to the change in trade policies, are more difficult to project. Only some of these effects can be simulated sensibly across countries.

Developing countries rely on import duties to a much larger extent than industrial countries. In African least developed countries (LDCs), import duties represented about 34 percent of total government revenue over the period 1999–2001, and exceeded 50 percent in a number of countries. In case trade taxes are included in the base of domestic taxes on imports, the reduction in trade taxes is usually accompanied by a reduction in VAT revenue, and it may also lead to a reduction in excises.

This paper contributes to the discussion on the revenue implications of trade reform by assessing the approximate fiscal revenue impact of liberalization formulae under consideration in the WTO for 58 low- and middle-income countries. Section II provides background information on trade negotiations under the GATT, the Uruguay Round, and the Doha Round. The relevant literature is summarized in Section III. Methodological aspects and the simulation results are presented in Section IV, as are country experiences with the rebalancing of fiscal revenue loss resulting from trade liberalization. The policy recommendations in Section V address the countries most affected by a decline in revenues.

II. Background

A. Negotiations under the GATT and the Uruguay Round

Over the past 25 years, trade liberalization has resulted in a sharp decline in the overall importance of revenue derived from customs duties and trade taxes. Reflecting commitments under trade liberalization agreements, as well as unilateral decisions, the collected import tariff rate10 fell by almost half since the mid-1980s. The trend has been most pronounced in low-income countries; however, even middle- and high-income countries experienced a sharp decline in the share of trade-derived revenue to GDP (IMF, 2005).

Since 1947, seven negotiation rounds under the GATT resulted in significant tariff reductions, although not covering trade in agricultural goods. Initially, the tariff negotiations under the GATT followed the request-and-offer procedure, under which members negotiated bilateral market access concessions which were subsequently extended to all members according to the MFN principle. Under the Kennedy Round (1963–67), a linear formula approach was introduced, resulting in a 50 percent cut of bound tariffs on all manufactured goods with the exception of “sensitive” goods, such as steel, clothing, textiles and footwear. Moreover, the negotiation agenda was extended beyond tariffs to anti-dumping measures. With the Tokyo Round (1973–79), negotiations on non-tariff measures11 gained importance.

Under the Uruguay Round (1987–94), all original GATT articles were reviewed and the WTO was established. The negotiation agenda was significantly expanded and covered trade in agriculture, textiles and apparel, and services.12 Key outcomes included the replacement of non-tariff barriers with bound tariffs, and the elimination—after a transition period—of quotas in textiles trade. The Agreement on Agriculture comprises specific binding commitments to improve market access and to reduce production-and trade-distorting domestic support and export subsidies; it also introduces a tariff rate quota13 and special safeguards provisions.14 However, despite these results, agricultural tariffs remained high15 and complex, with a considerable dispersion, and tariff escalation prevails in important product chains. Agreements on “Special and Differential Treatment” were reached providing longer implementation periods and lower reduction commitments for developing countries.

The later rounds of trade negotiations demonstrated the superiority of a formula-based approach that limits the role played by special interest groups and enables the effective participation of smaller countries that would not be able to conduct bilateral negotiations effectively. The 35 percent reduction in average tariffs resulting from the Kennedy Round, based on a 50 percent proportional formula, compares favorably with the average reduction of tariffs by 2.5 percent in the second through the fifth rounds of GATT negotiations under the request-and-offer approach (Francois and Martin, 2003).

Negotiations under the Uruguay Round brought about substantial tariff reductions, based on broad goals, such as a 36 percent average reduction of tariffs on industrial products. However, the round was less successful in reducing tariff dispersion as it left the distribution of the cuts across sectors to negotiations between trading partners. Formulae that would have harmonized tariffs in addition to reducing averages were proposed but not adopted—such as the Swiss formula16 that had originally been put forward by Switzerland in the Tokyo Round negotiations. The Swiss formula narrows the range of final tariff rates from a wide set of initial tariffs by applying steeper cuts to higher tariffs, while fixing a maximum final rate. It maintains the simplicity of a linear formula, requiring negotiations over only one coefficient, while reducing higher tariffs by more in both absolute and relative terms.

B. The Doha Round

Negotiations on the Doha Development Agenda began in November 2001 with the aim to agree on liberalization and rules in agricultural, industrial and services trade, with special consideration of the needs of developing countries. First substantive results were reached at a WTO General Council meeting in Geneva in July 2004, which adopted framework agreements for establishing modalities in the different negotiating areas (“July Framework”),17 including, crucially, agreements on the reduction of tariffs and domestic (agricultural) subsidies using harmonizing formulae, and the discontinuation of export subsidies.

At the sixth Ministerial Conference of the WTO in Hong Kong SAR in December 2005, trade ministers reached agreement on several outstanding issues but did not converge fully on negotiating modalities. The text on market access in agriculture goes slightly beyond the July Framework in adopting the principle of four tariff bands subject to progressively higher cuts, but fails to specify the thresholds of the bands and the size of the cuts. Absolute tariff caps were proposed by some (EU, United States, and G-2018) but rejected by others (G-1019). It was agreed that “sensitive products” can be excluded from formula cuts—though there would still need to be greater effective market access—and that developing countries can also make use of “special product” designations and a “special safeguard mechanism.”20 Regarding non-agricultural market access (NAMA), the Ministerial adopted a Swiss formula approach while reaffirming less than full reciprocity and special flexibilities for developing countries. The Ministerial also made progress in some other areas, e.g., setting a 2013 deadline for eliminating agricultural export subsidies and reaching agreement on duty-and-quota-free access to industrial country markets for products from LDCs. In a statement relevant to this paper, the final declaration also calls for greater clarity on the scope of the problem of tariff revenue dependency.21

III. Overview of the Literature

The literature provides a range of assessments of the likely overall effect of trade liberalization on revenue. Ebrill, Stotsky, and Gropp (1999) and Agbeyegbe, Stotsky, and WoldeMariam (2004) conclude that the revenue impact of trade liberalization for a sample of selected developing and emerging market economies tends to be limited. Overall, revenue tends to be least affected if the initial position of the trade regime is highly restrictive and if liberalization is accompanied by reforms in customs and tax administration, also with the aim to broaden the tax base. The revenue impact is reduced if measures involve the tariffication of quantitative restrictions; the auctioning of licenses; a reduction in tariff dispersion; and the elimination of exemptions. (Ebrill, Stotsky, and Gropp 1999)

By contrast, Khattry and Rao (2002) find that the impact of trade liberalization on revenue is significantly negative. For the period 1978–99, they posit that trade liberalization in developing countries resulted in revenue losses due to the inability to compensate for foregone tariff revenue by raising higher revenue from domestic sources.

Fisman and Wei (2004) analyze the relationship between tariff rates and evasion based on export data from Hong Kong SAR and import data from China, finding a fall in tariff evasion of 3 percent for each percentage point of reduction in the average tariff rate. On a similar note, Pritchett and Sethi (1993) find a nonlinear relationship between statutory tariff rates and collected rates, indicating that incentives for misreporting increase with higher tariff rates.

An OECD analysis of the impact of trade liberalization on 24 developing economies reaches the conclusion that countries with higher and more dispersed tariff barriers, while being well positioned to benefit from a tariff reform, are also more vulnerable to revenue loss. Depending on the initial levels of tariffs and binding overhangs,22 the trade, welfare and revenue impact of tariff reductions were found to differ considerably; while countries with higher initial tariffs and a lower binding overhang record a larger revenue loss, their welfare gains derived from trade creation are likely more substantial. Simulations of tariff cuts based on a Swiss formula with a coefficient of 10 point to a strong negative correlation between the trade and the revenue effects, i.e., the countries affected the most by revenue loss also experience the most significant trade creation. (OECD, 2004).

Analyzing a sample of 66 developing countries during 1974–98, Combes and Saadi-Sedik (2006) find that policy—induced trade openness results in an overall strengthening of government revenue.

IV. Simulation of the Fiscal Revenue Impact of Tariff Liberalization

A. Methodology

The study simulates the fiscal impact of different trade liberalization formulae under the Doha Round for a sample of 58 developing countries. The sample includes all low- and middle-income WTO member countries, for which data for the bound tariff, applied tariff, and imports for 2001 or later years are available at the requisite level of detail.23 The classification of countries by income and region is shown in Table 2. For the purpose of this study, data from the WITS database24 at the HS-6 digit level (5,113 tariff lines) were used. The simulations are limited to first round effects and do thus not take into account any possible compensation for revenue losses through second round effects. Section D provides a quantitative perspective on the possible impact of higher GDP growth resulting from trade liberalization on fiscal revenue. Moreover, the simulations do not consider existing tariff exemptions that would limit the impact of tariff cuts.25 Accordingly, they are testing the maximum loss in fiscal revenue based on statutory rates.

Table 2.

Classification of Sample Countries by Income and Region

article image
Source: IMF staff.

Following the initial analysis of data adequacy, several adjustments were made:

  • For 19 percent of tariff lines, on average, data for bound tariffs were not available, but data on applied tariffs existed. In these cases, mark-ups of 5 and 30 percentage points (to test sensitivity) were added to applied rates in order to construct an artificial “bound” tariff, in line with proposals reviewed in Annex B to the Hong Kong Ministerial Declaration.

  • For 0.7 percent of tariff lines, bound tariffs were available, while applied tariffs were not. In these cases, the applied tariff was set according to the ratio of the average not-missing bound and applied tariff.

  • 0.8 percent of tariff lines had neither bound nor applied tariffs. In these cases the applied rate was set to zero, and the bound rate was set with the 5 and 30 percentage point mark-up.

Formula reductions were made to bound tariffs at the HS-6 digit level of classification. In the simulations, the reduced bound tariffs were compared to the current statutory applied MFN tariff, and the lower of the two was used to calculate the projected revenue. Tariff cuts for agricultural goods were simulated for different sets of import demand elasticities, based on tariff allocation into four bands with progressively higher reduction coefficients, and applying different levels of tariff capping and sensitive product exclusions. It was assumed that countries would classify as sensitive products those tariff lines that carry most revenues (i.e., the selection criterion would be strictly fiscal, not reflecting trade policy concerns).

NAMA simulations were also run for different sets of import demand elasticities, by applying Swiss formula cuts and allowing for flexibilities for developing countries in line with Annex B, paragraph 8, of the July Framework. This paragraph grants the right (i) to exclude from formula reductions a maximum of 5 percent of tariff lines up to a total not exceeding 5 percent of imports; or (ii) to apply only half the formula cut to 10 percent of tariff lines not exceeding 10 percent of imports. Calculating revenue loss for the “5/5” and the “10/10” scenarios required identifying commodity lines that countries would choose to exclude in order to minimize the revenue loss. For this purpose, we computed the hypothetical revenue loss for each HS-6 commodity line, assuming full tariff cuts. We chose lines to be exempted from tariff cuts by minimizing the sum of revenue loss across non—exempt commodity lines subject to the dual constraints of paragraph 8 (a standard integer optimization problem), using the CPLEX solver integrated into GAMS (General Algebraic Modeling System).

B. Structure of tariff revenue in the sample

The analysis of initial conditions demonstrates that revenue is concentrated in tariff lines for non-agricultural goods, which account for an average of 78 percent of tariff revenue, largely because of a considerably higher import share in these categories of goods. However, for Bangladesh, tariff revenue from imports of agricultural products accounts for 59 percent of total tariff revenue, while it is equivalent to almost half the total tariff revenue for Papua New Guinea and Senegal. For other countries, such as Argentina, Brazil, Chile, Indonesia, Malaysia, and South Africa, the share of tariff revenue on agricultural imports is in the single digits. However, on average, agricultural tariffs are considerably higher than tariffs on non-agricultural products (Appendix Table 2). While the simple (trade weighted) average applied MFN tariff for non-agricultural products amounts to 9.8 (9.2) percent, it is equal to 16.5 (16.1) percent for agricultural products (Appendix Table 3).

Average bound tariffs remain substantially higher than applied tariffs, at 57.8 percent for agricultural products and 34.0 percent for non-agricultural products (Appendix Table 3). The “water” (i.e., margin) between bound and applied rates varies greatly between countries, as depicted in Figure 1 (see also Appendix Table 4). Since formula reductions are made to bound rates, this implies that the same reduction coefficient can have a very different impact on different countries.

Figure 1.
Figure 1.

Average Bound, Applied, and Collected Tariff Rates

Citation: IMF Working Papers 2006, 203; 10.5089/9781451864632.001.A001

Source: IMF staff.

Effectively collected tariffs, in turn, are generally lower than applied tariffs. Due to exemptions and tariff preferences, collected tariff rates amount to only 5.9 percent on average, compared to the trade-weighted average MFN tariff of 10.1 percent (for all products and countries; Appendix Table 3). For several countries (including China, Croatia, E1 Salvador, Honduras, Indonesia, Jamaica, Kyrgyz Republic, Malaysia, Oman, and Uganda), collected tariffs equal less than half of trade-weighted average MFN tariffs. In the cases of Georgia and Guyana, this ratio is less than one third, while it is as low as 20 percent for Madagascar, Mexico, and Turkey. While it is difficult to obtain detailed country-specific information on exemptions, this should be considered when interpreting the results of this study; specifically, the presence of exemptions reduces the fiscal impact of tariff reduction compared with statutory losses, and cutting exemptions offers opportunities for mitigating the decline in revenue.

The concentration of tariff revenue in the tariff lines with the highest protection is significant (Appendix Tables 5 and 6). As shown in Appendix Table 5, the 0.5 percent of tariff lines with the highest protection account for 45 percent of total tariff revenue on average, ranging from 23 percent for Uruguay to 71 percent for the Kyrgyz Republic. The 1 and 2 percent of tariff lines with the highest protection account for 54 percent and 66 percent on average, respectively. Tariff revenue derived from specific categories of products such as cars, tobacco, and alcohol, amounts to 0.1 to 0.2 percent of GDP on average, but is significantly higher in individual cases (Appendix Table 7). Similarly, while imports of oil account for 12.5 percent of total imports on average, tariff revenue from oil imports amounts to only 0.3 percent of GDP. However, such revenue is significant for selected countries, such as Malaysia (2.4 percent), Jamaica (2.0), Zimbabwe (2.1), and Guyana (1.8).

C. Results

We simulated the revenue impact of trade liberalization based on five pairs of market access formulae for agriculture and NAMA, as proposed under the Doha Round. Simulations for four of these pairs are summarized in Box 1, but would appear to be less relevant at the current stage of the negotiations.

The following detailed analysis is based on the formula structures outlined in the July Framework (WTO document WT/L/579) and in the Hong Kong Ministerial Declaration (WT/MIN(05)/DEC).

  • For agriculture: The July package indicated that tariff reductions would be made through a tiered formula applied to bound rates and that “progressivity in tariff reductions will be achieved through deeper cuts in higher tariffs with flexibilities for sensitive products.” The July package also made clear that LDCs would be exempted from any reduction obligations. However, the July package left open to negotiation the “number of bands, the thresholds for defining the bands, and the type of tariff reduction in each band,” as well as “the role of a tariff cap in a tiered formula with distinct treatment for sensitive products.” The Hong Kong Ministerial Declaration clarified that four bands will be used for structuring tariff cuts, but WTO members were not able to agree on the relevant thresholds.

  • For industrial goods: The July package indicated that NAMA tariff lines would be cut using a non-linear formula applied to bound tariffs on a line-by-line basis. An exemption from tariff cuts was agreed for members with a binding coverage of non-agricultural tariff lines of less than 35 percent provided that they agree to bind these tariffs. Some flexibility was granted to developing countries in the form of the “5/5” and “10/10” provisions described in Section IV. A. above. Finally, as for agriculture, LDCs are not expected to cut their tariff rates. The Hong Kong Ministerial Declaration clarified that the non-linear formula would be a Swiss formula and that, for commodity lines that had not been bound previously, a non-linear mark-up approach would be used to derive bound rates, to which tariff reductions would be applied.

For coefficient values, we drew on the EU and G20 agriculture proposals of October 2005,26 namely:

  • 40 percent cut of bound rates higher than 130 percent;

  • 35 percent cut of bound rates between 80 and 130 percent;

  • 30 percent cut of bound rates between 30 and 80 percent;

  • 25 percent cut of bound rates lower than 30 percent.

For NAMA, we simulated results for four coefficients that fall in the range of Swiss formula coefficients (ranging from 10 to 25) currently under consideration.

Selected Tariff Reduction Scenarios1

1. Linear Uruguay Round formula: bound tariffs reduced by 36 percent

Some early proposals such as the 2000 EU proposal on agriculture (WTO document G/AG/NG/W/90) suggested to continue with the Uruguay Round approach, i.e., an average 36 percent cut in bound tariffs with a minimum 15 percent cut for each tariff line. Our first simulation applied a 36 percent tariff cut to all bound tariff lines (agricultural and non- agricultural), with and without a capping of agricultural tariff lines at 100 percent, and with and without exempting LDCs and countries that have bound less than 35 percent of industrial tariff lines. Tariff losses in this scenario would average 20 percent, most pronounced among low-middle-income countries, in particular in the Middle East and Central Asia and in Asia. However, most LDCs would face a decline in tariff revenue of less than 1 percent. The application of exemptions granted to countries with less than 35 percent of their industrial tariff lines bound would lead to an insignificant decline in tariff revenue losses on average.

2. U.S. proposals of 2002-03

Agricultural products: Swiss formula on applied tariff, coefficient 25

Industrial products: Swiss formula on bound tariff, coefficient 8; rates below 5 percent reduced to zero.

The implementation of this proposal (WTO document TN/MA/W/18) would result in substantial tariff revenue loss for all countries, with an average loss of 73 percent of revenue, and 70 percent of revenue in case exemptions from tariff cuts were applied. The impact for the 10 countries most exposed to revenue loss would be severe, as indicated by an average loss of tariff revenue of 82 percent, taking into account exemptions. Overall, losses would be concentrated among the lower middle-income group, while exemptions would reduce the revenue loss experienced by African countries.

3. Combination of Harbinson proposal for agriculture with Swiss formula for NAMA

article image

Industrial products: Swiss formula on bound tariff, with a coefficient of 8

Under this proposal, all countries would experience a tariff revenue loss, with the average amounting to 59 percent of initial revenue. Losses in terms of GDP would average 2.3 percent, and low-income countries would experience slightly higher revenue losses than the other country groups.

4. Pakistan proposal

In July 2005, Pakistan made the proposal (WTO document TN/MA/W/60) to apply a single Swiss formula to NAMA bound tariffs using as a coefficient the average tariff rate (one for developed and one for developing countries). For developing countries the coefficient would be 29. The simulation added to this formula a 5 percent reduction in bound tariff of agricultural goods, in order to attain an estimate of the entire tariff cut. In this scenario, tariff revenue loss, experienced by all countries, would average 30 percent of revenue, the equivalent of 1.2 percent of GDP. The revenue decline would vary widely by income group, between 22 percent of revenue for higher middle-income countries and 38 percent for low-income countries.

1/ Based on data for 2001-04.


As noted, our simulations are based on the coefficients proposed by the EU and the G-20. In order to account for remaining differences on capping and flexibilities we simulated:

  • Four different levels of capping, i.e. maximum tariffs: capping at 75 percent (U.S. proposal), 100 percent, 150 percent (EU and G-20), and no capping (G-10 including Japan); and

  • Four levels of sensitive product flexibilities: 1 percent (U.S. proposal), 2 percent, 4 percent, and 8 percent (EU proposal) of tariff lines are excluded from tariff cuts. As noted previously, in our simulation “sensitive products” were those associated with the highest pre-liberalization fiscal revenue.27 Consistent with the July Framework and the Hong Kong Ministerial Declaration, tariffs are not cut for LDCs in the sample.

Results under a scenario with import demand elasticities of zero are presented in Table 3. Absent capping, the weighted average revenue loss derived from cuts in agricultural tariffs would amount to 23 percent, while the simple average revenue loss would be less than 9 percent (equivalent to 0.08 percent of GDP). The significant difference indicates that “big countries” with a large share of revenue derived from agricultural imports would stand to lose more than other countries. For instance, China and Mexico would lose 48 and 32 percent, respectively, of tariff revenue on agricultural imports. When exemptions for LDCs are taken into account, the low-income countries in the sample would see their tariff revenue decline by only 3 percent, while the decline in middle-income countries would be about one quarter. Capping would have a significant impact only if set at a level below 100 percent.

Table 3.

Summary of Results for Agricultural Tariff Cuts 1/

(Percent change in revenue from agricultural tariffs)

article image
Source: IMF staff estimates.

Assuming import demand elasticity of zero; exemptions for LDCs apply.

Simple average.

The assumption of an import demand elasticity of zero does not seem realistic as lower prices of imported goods following the tariff cut would trigger an increase in demand for these goods, thereby partially mitigating the revenue loss. Accordingly, country-and product-specific elasticities at the HS-6 digit level were introduced into the simulation, based on the estimates of Kee, Nicita, and Olarreaga (2004). The simulation results are summarized in Table 4. As expected, the introduction of elasticities would mitigate the revenue loss considerably. With the introduction of elasticities, the effect of capping is reduced significantly, since high tariffs can be virtually prohibitive and imports in these categories grow by more than the tariff declines.

Table 4.

Summary of Results for Agricultural Tariff Cuts 1/

(Percent change in revenue from agricultural tariffs)

article image
Source: IMF staff estimates.

Assuming import demand elasticities as in Kee et al. (2004); exemptions for LDCs apply.

Simple average.

Non-Agricultural Market Access (NAMA)

Simulations for NAMA were based on Swiss formula tariff cuts with different coefficients and take into account the flexibilities for developing countries described earlier.28 While proposals for coefficients range from 8 to 30 and one proposal (by Argentina, Brazil, and India) postulates country-specific coefficients, we chose coefficients of 10, 15, 20, and 25, which seem representative of the range of possible outcomes. We also allowed for flexibilities for developing countries in line with Annex B, paragraph 8 of the July Framework, employing the methodology set out above in Section IV. A (“5/5” and “10/10” flexibilities). The proposals were tested both with an import demand elasticity of zero and with a demand response based on the import demand elasticities calculated by Kee, Nicita, and Olarreaga (2004) (Appendix Tables 8 and 9). However, unlike for agriculture, the choice of elasticities was shown to have only a moderate effect on the results, and is not reported here.

The projected revenue loss as a share of GDP for Swiss cut coefficients of 15, 20, and 25 is shown in Table 5 (see Appendix Table 10 for the full set of country-specific results). The average cut in the statutory tariff revenue for all developing countries in the sample ranges from 0.3 percent of GDP for a Swiss formula coefficient of 25 and “10/10 flexibilities” to 0.8 percent of GDP assuming a coefficient of 15 and the absence of flexibilities. Average results are skewed upward by a few outliers in each income group: in the LDC-group, Maldives constitutes such an outlier with simulated fiscal revenue loss equivalent to 3½; percent of GDP even under the most benign of constellations. It is worth repeating, however, that according to the July Framework and the Hong Kong Ministerial Declaration, LDCs will not be expected to make formula cuts in their tariffs.

Table 5.

NAMA Products: Average Revenue Loss with Swiss Cut and Flexibilities

(In percent of GDP)

article image
Source: IMF staff estimates.

A scenario with a coefficient of 15 is depicted in Figure 2, ranking the 48 non-LDCs in the sample by the size of the revenue impact of formula cuts (the results are reported in more detail in Appendix Table 10). The impact of formula cuts is significant relative to base-year revenue from import tariffs, but for the majority of countries very small relative to GDP. (Appendix Table 11 shows country-specific revenue losses in terms of original tariff revenue.)

Figure 2.
Figure 2.

Loss of Tariff Revenue: NAMA Swiss Formula, 15 Coefficient

(In percent of GDP)

Citation: IMF Working Papers 2006, 203; 10.5089/9781451864632.001.A001

Source: IMF staff estimates.

The simulations suggest that under the base line scenario, without flexibilities, assuming a NAMA Swiss formula coefficient of 15, about three quarters of the countries under analysis would experience a decline in tariff revenue of 30 percent or less of initial revenue. In less than 10 percent of the cases would the revenue loss exceed 1 percent of GDP, especially in a scenario in which tariffs on oil import are “converted” into excises that would be excluded from reductions, since revenue from oil imports tends to be disproportionately important for the most revenue-sensitive countries. Still, for a minority of countries the revenue effects are clearly significant enough to warrant mitigating policies (see Figures 2 and 3).

Figure 3.
Figure 3.

Loss of Tariff Revenue (In percent of original tariff revenue

Citation: IMF Working Papers 2006, 203; 10.5089/9781451864632.001.A001

Source: IMF staff estimates.

D. Growth simulations

In order to assess the possibility of recouping revenue losses arising from trade liberalization measures, we carried out additional simulations. To gain a rough numerical understanding of the orders of magnitude involved, we analyzed the average loss of revenue resulting from tariff reductions and developed a scenario that would allow the country to regain the preliberalization level of revenues. The scenario is based on data for the “average” country and a set of assumptions as outlined below.29 For the representative country, the average applied tariff in the sample would fall from 10.7 percent to 8.2 percent, and the tax revenue-to-GDP ratio would fall from a pre-liberalization level of 19.5 percent to a post-liberalization level of 18.7 percent, resulting in fiscal revenue losses equivalent to ¾ of a percent of GDP.

Table 6 presents the number of years it will take to regain the pre-liberalization ratio of tax revenue-to-GDP following the implementation of trade liberalization measures, under the assumption of a constant pre-liberalization nominal annual GDP growth rate of 4 percent. The table shows results for various assumptions regarding tax elasticity (εT), income elasticity of imports (εI), and additional annual GDP growth pick-up (Δg) associated with trade liberalization.

Table 6.

Years Needed to Reach Pre-Liberalization Tax Revenue-to-GDP Ratio

article image
Source: IMF staff estimates.

As could be expected, the period of time needed to fully regain the pre-liberalization tax revenue-to-GDP ratio is the longest—10 years—in case no additional growth is assumed (Δg = 0), and the income elasticity of imports (εI) as well as the tax elasticity (εT) are relatively low at 1.1. Increasing additional annual GDP growth to 0.75, tax elasticity to 1.3 and income elasticity of imports to 1.5 shortens the period of time needed to regain the pre-liberalization ratio of revenue to GDP to 2-3 years. It should be noted that the simulations do not imply a recovery of losses accumulated during the interim period.

V. Summary and Policy Recommendations

This paper responds to the declaration of the Hong Kong SAR Ministerial meeting calling for more analysis of the scope of the tariff dependency problem in the context of the Doha Development Agenda. It also constitutes an input to the Fund’s bilateral and regional surveillance work, and indicates a potential need for assistance to selected member countries in embarking on trade reform.

Experiences with Compensatory Tax Reform

The experiences of different countries with the effects of tax reforms aimed at mitigating revenue loss from trade liberalization differ. Moreover, the assessment of these experiences also varies, depending on the methodology and the data sources used. Khattry and Rao (2002) underline the difficulties developing countries face in identifying and implementing adequate compensatory revenue measures. By contrast, Keen and Simone (2004) posit that revenue recovery has been almost complete.

In a detailed study, Baunsgaard and Keen (2005) analyze the implications of trade liberalization for revenue for 125 countries between 1975 and 2000. The results show substantial variations by income group:

  • For low-income countries, revenue recovery remained low, at about 30 percent of lost trade tax revenue,1 the result being independent of the application of a VAT over the short term. Non-linearity implied that recovery became less likely if the initial level of collected trade taxes was low.

  • Middle-income countries showed a somewhat stronger record, with recovery ratios reaching 45 to 65 percent.

  • By contrast, high-income countries demonstrated a recovery of lost trade taxes through domestic taxes of more than 100 percent, with the highest ratios reached by countries relying on VAT.2

Revenue recovery was reported to have been lower in the second half of the period under investigation. By region, recovery rates were the lowest in Asia and the Middle East. However, country experiences were found to have been rather diverse across regions.3

In several countries, trade liberalization was implemented in parallel with the introduction of the VAT, thereby achieving revenue neutrality. In this regard, country experiences point to the importance of a strong commitment to broadening the tax base and proper design and implementation of the VAT. Experience suggests that VAT systems complying with standard best practice of a single base, a reasonable threshold, and few exceptions, were best geared toward offsetting lost trade tax revenue.4

1/ However, revenue recovery has been strong in some low-income countries, such as Benin, Cote d’Ivoire, Gambia, Malawi, Pakistan, and Zambia.2/ The authors conclude that for this group of countries, with regard to trade taxes, protective motives dominated over revenue considerations.3/ See also Greenaway and Milner (1991), and IMF (2005) for detailed case studies.4/ Practical problems may, however, arise as domestic VAT collection faces stronger implementation difficulties than the collection of VAT on imports (the latter accounting for the major part of all VAT revenue in many developing countries).

The simulations of key proposals currently under discussion in the context of the Doha Round indicate that, under the baseline scenario (see above), only 10 percent of developing countries would face a loss in tariff revenue exceeding 1 percent of GDP. As agricultural imports account for only a small share of tariff revenue and the formulae under discussion would produce only moderate reductions in applied tariffs, the results are largely determined by NAMA liberalization.

Our simulations of multilateral trade liberalization measures suggest that most countries would face only minor fiscal pressures, of a magnitude that is likely to be mitigated by second-round effects. But a few countries may have to consider complementary reforms, such as the reform of non-tariff barriers, the reduction in exemptions, and perhaps a broader shift of taxation toward domestic taxes (see Box 2 for country experiences with tax reform measures, following trade liberalization). As second-round effects are subject to significant uncertainty as to their strength and timing, it is generally not prudent to presume that effects through exchange rate movements, inflation, or economic growth would automatically and immediately compensate for the direct loss of trade tax revenues.

A combination of measures to broaden the tax base, strengthen tax administration, and reduce exemptions could achieve the intended revenue compensation. If domestic indirect tax rates are raised to a level that leaves end prices for consumers unchanged, there would be efficiency gains from aligning domestic producer prices closer with world market prices.30 Moreover, government revenue would likely increase as the domestic tax would be imposed on the consumption of domestic as well as imported goods.31 Vice versa, the reform could aim to compensate the revenue loss exactly, in which case indirect tax rates and thus consumer prices could be reduced. Welfare gains could be maximized by combining the shift away from trade taxes with a move toward a uniform system of consumption taxes (IMF, 2005).

Additional reductions in impact could be attained by transforming some import tariffs into excises, thereby excluding them from tariff reductions. The collection of both sales tax and excise could be based on the administrative infrastructure developed for trade tax collection, as a significant share would continue to be imposed on imports and collected at the border. Adding to the attractiveness of the proposed shift from a practical point of view, the sales tax could be expected to cover the informal sector in the same way as tariffs do.

As efficiency gains from trade liberalization would likely lead to an increase in aggregate real incomes, additional government revenue could be derived from higher income taxes (Blejer and Cheasty, 1990; IMF, 2005). Overall, policies should be based on mutual reinforcement of trade reform, domestic tax reform, and sound macro policies.

Fiscal Implications of Multilateral Tariff Cuts
Author: Mr. Azim M Sadikov, Mr. Hans P Lankes, Mr. Dustin Smith, Ms. Katrin Elborgh-Woytek, and Mr. Jean-Jacques Hallaert
  • View in gallery

    Average Bound, Applied, and Collected Tariff Rates

  • View in gallery

    Loss of Tariff Revenue: NAMA Swiss Formula, 15 Coefficient

    (In percent of GDP)

  • View in gallery

    Loss of Tariff Revenue (In percent of original tariff revenue